Disappointment under rescue rule S3, k=2. generalisation check, gap 0.422. 84.3 % of clips score at or below zero on this emotion and the largest gap on its normalised axis is 0.422 (WIDER than the 0.25 step cap). This rule found 13,495 chains over 40,000 tracks; the strict rule found 0 at k=3.
S3 changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'. What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained. Full explanation →Interest has a median of 2.08 and is never zero, while Infatuation is zero on 87.7 % of clips — and the caption named an emotion whenever its raw score cleared an absolute 1.0. Interest therefore appeared in 94.8 % of captions and Sadness in almost none: the clause was reporting the scale of the head, not the emotion of the clip. An emotion is now named only when it lands in the top 10 % for that emotion, against a pooled tie-aware mid-rank ECDF over 132,833,726 utterances spanning every dataset and language. Interest now appears in 6.2 %, all 40 emotions occur, and a clip that is ordinary on all 40 says “no dominant emotion” rather than being forced to pick one (21.8 % of clips). This is the same scale the trajectory miner selects on, so the caption and the mining now refer to the same quantity: the mined target emotion is named in the final clip's caption on 73 % of chains, up from 46 %.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.01. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.01.
On the corpus-wide percentile scale those become 0.39, 0.94 — a total move of +0.54.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.721 before conversion and 0.794 after — it rose by 0.073. Neighbour-to-neighbour the worst pair went 0.721 → 0.794. (The earlier render, with segment 1 left raw, scores 0.704 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.543 in the original and +0.525 after conversion — 97 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.93 → 3.15 (+0.22) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.26. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.26.
On the corpus-wide percentile scale those become 0.39, 0.98 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.727 before conversion and 0.770 after — it rose by 0.044. Neighbour-to-neighbour the worst pair went 0.727 → 0.770. (The earlier render, with segment 1 left raw, scores 0.742 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.587 in the original and +0.597 after conversion — 102 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.86 → 3.13 (+0.27) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.30. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.30.
On the corpus-wide percentile scale those become 0.39, 0.99 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.793 before conversion and 0.797 after — it rose by 0.004. Neighbour-to-neighbour the worst pair went 0.793 → 0.797. (The earlier render, with segment 1 left raw, scores 0.719 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.591 in the original and +0.534 after conversion — 90 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 3.04 → 3.13 (+0.09) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.06. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.06.
On the corpus-wide percentile scale those become 0.39, 0.95 — a total move of +0.56.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.020 before conversion and 0.727 after — it rose by 0.707. Neighbour-to-neighbour the worst pair went 0.020 → 0.727. (The earlier render, with segment 1 left raw, scores 0.642 here.) This chain started below 0.50 — the segments really were different people, the band the conversion helps most: chains starting below 0.50 improve on this measure almost without exception.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.557 in the original and +0.004 after conversion — 1 % of the delta retained, so a meaningful part of the trajectory was flattened.
Quality. Mean predicted overall quality across the segments went 2.82 → 3.14 (+0.31) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.28. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.28.
On the corpus-wide percentile scale those become 0.39, 0.98 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.359 before conversion and 0.516 after — it rose by 0.157. Neighbour-to-neighbour the worst pair went 0.359 → 0.516. (The earlier render, with segment 1 left raw, scores 0.398 here.) This chain started below 0.50 — the segments really were different people, the band the conversion helps most: chains starting below 0.50 improve on this measure almost without exception.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.589 in the original and +0.447 after conversion — 76 % of the delta retained, which is most of it.
Quality. Mean predicted overall quality across the segments went 2.98 → 3.24 (+0.26) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.29. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.29.
On the corpus-wide percentile scale those become 0.39, 0.98 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.916 before conversion and 0.917 after — it rose by 0.000. Neighbour-to-neighbour the worst pair went 0.916 → 0.917. (The earlier render, with segment 1 left raw, scores 0.804 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.590 in the original and +0.162 after conversion — 27 % of the delta retained, so a meaningful part of the trajectory was flattened.
Quality. Mean predicted overall quality across the segments went 2.93 → 3.10 (+0.17) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.11. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.11.
On the corpus-wide percentile scale those become 0.39, 0.96 — a total move of +0.57.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.434 before conversion and 0.770 after — it rose by 0.336. Neighbour-to-neighbour the worst pair went 0.434 → 0.770. (The earlier render, with segment 1 left raw, scores 0.607 here.) This chain started below 0.50 — the segments really were different people, the band the conversion helps most: chains starting below 0.50 improve on this measure almost without exception.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.567 in the original and +0.528 after conversion — 93 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.85 → 3.00 (+0.15) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.17. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.17.
On the corpus-wide percentile scale those become 0.39, 0.97 — a total move of +0.58.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 100–100 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.907 before conversion and 0.906 after — it fell by 0.001. Neighbour-to-neighbour the worst pair went 0.907 → 0.906. (The earlier render, with segment 1 left raw, scores 0.814 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.577 in the original and +0.553 after conversion — 96 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.84 → 3.21 (+0.37) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.20. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.20.
On the corpus-wide percentile scale those become 0.39, 0.98 — a total move of +0.58.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.710 before conversion and 0.723 after — it rose by 0.012. Neighbour-to-neighbour the worst pair went 0.710 → 0.723. (The earlier render, with segment 1 left raw, scores 0.606 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.581 in the original and +0.496 after conversion — 85 % of the delta retained, which is most of it.
Quality. Mean predicted overall quality across the segments went 2.61 → 2.94 (+0.33) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.35. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.35.
On the corpus-wide percentile scale those become 0.39, 0.99 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.902 before conversion and 0.901 after — it fell by 0.001. Neighbour-to-neighbour the worst pair went 0.902 → 0.901. (The earlier render, with segment 1 left raw, scores 0.840 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.594 in the original and +0.596 after conversion — 100 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.94 → 3.25 (+0.31) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.22. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.22.
On the corpus-wide percentile scale those become 0.39, 0.98 — a total move of +0.58.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.802 before conversion and 0.906 after — it rose by 0.104. Neighbour-to-neighbour the worst pair went 0.802 → 0.906. (The earlier render, with segment 1 left raw, scores 0.776 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.583 in the original and +0.000 after conversion — 0 % of the delta retained, so a meaningful part of the trajectory was flattened.
Quality. Mean predicted overall quality across the segments went 2.99 → 3.24 (+0.25) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.18. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.18.
On the corpus-wide percentile scale those become 0.39, 0.97 — a total move of +0.58.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.872 before conversion and 0.900 after — it rose by 0.027. Neighbour-to-neighbour the worst pair went 0.872 → 0.900. (The earlier render, with segment 1 left raw, scores 0.879 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.578 in the original and +0.537 after conversion — 93 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.83 → 3.22 (+0.39) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.24. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.24.
On the corpus-wide percentile scale those become 0.39, 0.98 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.851 before conversion and 0.890 after — it rose by 0.039. Neighbour-to-neighbour the worst pair went 0.851 → 0.890. (The earlier render, with segment 1 left raw, scores 0.776 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.586 in the original and +0.485 after conversion — 83 % of the delta retained, which is most of it.
Quality. Mean predicted overall quality across the segments went 3.03 → 3.25 (+0.22) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.33. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.33.
On the corpus-wide percentile scale those become 0.39, 0.99 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 100–100 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.897 before conversion and 0.899 after — it rose by 0.002. Neighbour-to-neighbour the worst pair went 0.897 → 0.899. (The earlier render, with segment 1 left raw, scores 0.787 here.) This chain started at or above 0.80 — already effectively one voice. Across the whole build that is the band where conversion tends to cost identity agreement rather than add it, and a hard identity cut would keep this chain without converting it at all. Judge it by ear against the original above.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.593 in the original and +0.570 after conversion — 96 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.82 → 3.10 (+0.29) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.01, 1.05. In the first clip the scorer found no Disappointment whatsoever (0.01); by the last it is at 1.05.
On the corpus-wide percentile scale those become 0.79, 0.95 — a total move of +0.16.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 100–100 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.761 before conversion and 0.840 after — it rose by 0.078. Neighbour-to-neighbour the worst pair went 0.761 → 0.840. (The earlier render, with segment 1 left raw, scores 0.803 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.158 in the original and +0.535 after conversion — 339 % of the delta retained, i.e. the move came out slightly larger after conversion than before.
Quality. Mean predicted overall quality across the segments went 2.56 → 2.94 (+0.38) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.38. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.38.
On the corpus-wide percentile scale those become 0.39, 0.99 — a total move of +0.60.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was -0.024 before conversion and 0.942 after — it rose by 0.966. Neighbour-to-neighbour the worst pair went -0.024 → 0.942. (The earlier render, with segment 1 left raw, scores 0.792 here.) This chain started below 0.50 — the segments really were different people, the band the conversion helps most: chains starting below 0.50 improve on this measure almost without exception.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.595 in the original and +0.587 after conversion — 99 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 3.14 → 3.33 (+0.20) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.18. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.18.
On the corpus-wide percentile scale those become 0.39, 0.97 — a total move of +0.58.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.735 before conversion and 0.771 after — it rose by 0.036. Neighbour-to-neighbour the worst pair went 0.735 → 0.771. (The earlier render, with segment 1 left raw, scores 0.750 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.579 in the original and +0.521 after conversion — 90 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.76 → 2.99 (+0.23) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is -0.00, 1.10. In the first clip the scorer found no Disappointment whatsoever (-0.00); by the last it is at 1.10.
On the corpus-wide percentile scale those become 0.39, 0.96 — a total move of +0.57.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 100–100 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.281 before conversion and 0.642 after — it rose by 0.362. Neighbour-to-neighbour the worst pair went 0.281 → 0.642. (The earlier render, with segment 1 left raw, scores 0.542 here.) This chain started below 0.50 — the segments really were different people, the band the conversion helps most: chains starting below 0.50 improve on this measure almost without exception.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.960 in the original and +0.940 after conversion — 98 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.74 → 3.13 (+0.38) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is 0.00, 1.08. In the first clip the scorer found no Disappointment whatsoever (0.00); by the last it is at 1.08.
On the corpus-wide percentile scale those become 0.39, 0.96 — a total move of +0.56.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 100–100 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.746 before conversion and 0.710 after — it fell by 0.037. Neighbour-to-neighbour the worst pair went 0.746 → 0.710. (The earlier render, with segment 1 left raw, scores 0.731 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.561 in the original and +0.000 after conversion — 0 % of the delta retained, so a meaningful part of the trajectory was flattened.
Quality. Mean predicted overall quality across the segments went 2.60 → 2.97 (+0.38) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.
This is not a strict-rule trajectory. It comes from rescue rule S3, which exists because the strict rule returns nothing at all for Disappointment.
The raw scorer output across the chain is -0.00, 1.31. In the first clip the scorer found no Disappointment whatsoever (-0.00); by the last it is at 1.31.
On the corpus-wide percentile scale those become 0.39, 0.99 — a total move of +0.59.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Disappointment, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Disappointment sit above 0.85. There is simply nothing in between to step onto, so the only move available is one jump far wider than the 0.25 per-step cap.
What this rule changes: Absent to present. The chain must start at or below 0.05 (the emotion is absent) and end at or above 1.0 (it is clearly present). The most literal reading of 'from not sad to sad'.
What it costs: Says nothing about the SHAPE of the path -- only that it begins absent and ends present. Any intermediate clip is unconstrained.
Same speaker? No similarity score is available here — the emolia clips in this chain are not covered by either speaker-embedding store. The chain therefore rests on the corpus's own speaker/track labelling, which is not the same as a measured check.
Voice consistency: these clips are separate recordings joined together, and no voice-similarity check could be run for this sample, so there is no measurement of how closely the voices match. You may hear the voice shift between segments. Voice conversion has not been applied yet in this build. A planned pass will re-render every segment onto the first segment's voice, which removes this effect entirely.
What was done to this chain. All 2 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.
Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.775 before conversion and 0.862 after — it rose by 0.087. Neighbour-to-neighbour the worst pair went 0.775 → 0.862. (The earlier render, with segment 1 left raw, scores 0.776 here.) This chain started 0.70-0.80 — close, but under the identity threshold, a band where the conversion is close to a wash on this measure.
Did the emotion survive? Re-scored end to end through the same emotion stack and the same corpus-percentile scale the chain was mined on, Disappointment moved +0.986 in the original and +0.940 after conversion — 95 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 3.00 → 3.20 (+0.19) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.