Sadness under rescue rule S3, k=3. the same, with one unconstrained clip in between. 90.0 % of clips score at or below zero on this emotion and the largest gap on its normalised axis is 0.449 (WIDER than the 0.25 step cap). This rule found 5,806 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 Sadness.
The raw scorer output across the chain is 0.00, 0.60, 1.09. In the first clip the scorer found no Sadness whatsoever (0.00); by the last it is at 1.09.
On the corpus-wide percentile scale those become 0.43, 0.94, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 0.748 before conversion and 0.635 after — it fell by 0.113. Neighbour-to-neighbour the worst pair went 0.744 → 0.642. (The earlier render, with segment 1 left raw, scores 0.637 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, Sadness moved +0.551 in the original and +0.530 after conversion — 96 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.85 → 3.05 (+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 Sadness.
The raw scorer output across the chain is 0.00, 0.57, 1.19. In the first clip the scorer found no Sadness whatsoever (0.00); by the last it is at 1.19.
On the corpus-wide percentile scale those become 0.43, 0.93, 0.99 — 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 Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.304 before conversion and 0.407 after — it rose by 0.103. Neighbour-to-neighbour the worst pair went 0.059 → 0.407. (The earlier render, with segment 1 left raw, scores 0.380 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, Sadness moved +0.556 in the original and +0.539 after conversion — 97 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.78 → 3.20 (+0.42) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.67, 1.16. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.16.
On the corpus-wide percentile scale those become 0.43, 0.94, 0.99 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.173 before conversion and 0.663 after — it rose by 0.490. Neighbour-to-neighbour the worst pair went 0.256 → 0.684. (The earlier render, with segment 1 left raw, scores 0.487 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, Sadness moved +0.555 in the original and +0.549 after conversion — 99 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.73 → 3.04 (+0.32) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.02, 1.02. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.02.
On the corpus-wide percentile scale those become 0.43, 0.87, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.767 before conversion and 0.787 after — it rose by 0.020. Neighbour-to-neighbour the worst pair went 0.610 → 0.618. (The earlier render, with segment 1 left raw, scores 0.622 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, Sadness moved +0.548 in the original and +0.543 after conversion — 99 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 3.06 → 3.17 (+0.11) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.56, 1.09. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.09.
On the corpus-wide percentile scale those become 0.43, 0.93, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.750 before conversion and 0.791 after — it rose by 0.042. Neighbour-to-neighbour the worst pair went 0.766 → 0.791. (The earlier render, with segment 1 left raw, scores 0.806 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, Sadness moved +0.552 in the original and +0.516 after conversion — 93 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 3.08 → 3.23 (+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 Sadness.
The raw scorer output across the chain is 0.00, 0.36, 1.01. In the first clip the scorer found no Sadness whatsoever (0.00); by the last it is at 1.01.
On the corpus-wide percentile scale those become 0.43, 0.91, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 0.921 before conversion and 0.897 after — it fell by 0.024. Neighbour-to-neighbour the worst pair went 0.911 → 0.897. (The earlier render, with segment 1 left raw, scores 0.869 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, Sadness moved +0.547 in the original and +0.545 after conversion — 100 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.70 → 3.04 (+0.34) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.45, 1.30. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.30.
On the corpus-wide percentile scale those become 0.43, 0.92, 0.99 — 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 Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.777 before conversion and 0.773 after — it fell by 0.004. Neighbour-to-neighbour the worst pair went 0.777 → 0.773. (The earlier render, with segment 1 left raw, scores 0.690 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.
The emotional move did not survive. Re-scored end to end, Sadness moved +0.560 in the original and -0.019 after conversion — it changed direction. On this chain the corrected audio is not an improvement.
Quality. Mean predicted overall quality across the segments went 2.76 → 3.03 (+0.28) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.96, 1.67. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.67.
On the corpus-wide percentile scale those become 0.43, 0.97, 1.00 — 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 Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.867 before conversion and 0.787 after — it fell by 0.081. Neighbour-to-neighbour the worst pair went 0.867 → 0.787. (The earlier render, with segment 1 left raw, scores 0.751 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, Sadness moved +0.567 in the original and +0.562 after conversion — 99 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.78 → 2.99 (+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 Sadness.
The raw scorer output across the chain is 0.00, 0.64, 1.05. In the first clip the scorer found no Sadness whatsoever (0.00); by the last it is at 1.05.
On the corpus-wide percentile scale those become 0.43, 0.94, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.908 before conversion and 0.831 after — it fell by 0.077. Neighbour-to-neighbour the worst pair went 0.923 → 0.811. (The earlier render, with segment 1 left raw, scores 0.810 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, Sadness moved +0.549 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 3.00 → 3.15 (+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 Sadness.
The raw scorer output across the chain is -0.00, 0.48, 1.05. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.05.
On the corpus-wide percentile scale those become 0.43, 0.92, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.860 before conversion and 0.795 after — it fell by 0.066. Neighbour-to-neighbour the worst pair went 0.869 → 0.823. (The earlier render, with segment 1 left raw, scores 0.735 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, Sadness moved +0.550 in the original and +0.505 after conversion — 92 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.96 → 3.10 (+0.14) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.80, 1.20. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.20.
On the corpus-wide percentile scale those become 0.43, 0.96, 0.99 — 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 Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 0.086 before conversion and 0.422 after — it rose by 0.336. Neighbour-to-neighbour the worst pair went 0.201 → 0.427. (The earlier render, with segment 1 left raw, scores 0.396 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, Sadness moved +0.557 in the original and +0.476 after conversion — 85 % of the delta retained, which is most of it.
Quality. Mean predicted overall quality across the segments went 2.58 → 2.76 (+0.18) 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 Sadness.
The raw scorer output across the chain is 0.00, 0.97, 1.15. In the first clip the scorer found no Sadness whatsoever (0.00); by the last it is at 1.15.
On the corpus-wide percentile scale those become 0.43, 0.97, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 0.278 before conversion and 0.496 after — it rose by 0.219. Neighbour-to-neighbour the worst pair went 0.278 → 0.502. (The earlier render, with segment 1 left raw, scores 0.431 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, Sadness moved +0.555 in the original and +0.525 after conversion — 95 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.92 → 3.18 (+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 Sadness.
The raw scorer output across the chain is -0.00, 0.45, 1.05. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.05.
On the corpus-wide percentile scale those become 0.43, 0.92, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.334 before conversion and 0.776 after — it rose by 0.442. Neighbour-to-neighbour the worst pair went 0.334 → 0.776. (The earlier render, with segment 1 left raw, scores 0.616 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, Sadness moved +0.549 in the original and +0.531 after conversion — 97 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.80 → 3.08 (+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 Sadness.
The raw scorer output across the chain is -0.00, 0.44, 1.58. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.58.
On the corpus-wide percentile scale those become 0.43, 0.92, 1.00 — 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 Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.003 before conversion and 0.654 after — it rose by 0.651. Neighbour-to-neighbour the worst pair went 0.062 → 0.692. (The earlier render, with segment 1 left raw, scores 0.596 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, Sadness moved +0.566 in the original and +0.566 after conversion — 100 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.94 → 3.06 (+0.12) 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 Sadness.
The raw scorer output across the chain is 0.00, 0.67, 1.15. In the first clip the scorer found no Sadness whatsoever (0.00); by the last it is at 1.15.
On the corpus-wide percentile scale those become 0.43, 0.94, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.712 before conversion and 0.715 after — it rose by 0.003. Neighbour-to-neighbour the worst pair went 0.723 → 0.696. (The earlier render, with segment 1 left raw, scores 0.727 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, Sadness moved +0.555 in the original and +0.540 after conversion — 97 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.79 → 3.01 (+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 Sadness.
The raw scorer output across the chain is -0.00, 0.44, 1.11. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.11.
On the corpus-wide percentile scale those become 0.43, 0.92, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.834 before conversion and 0.865 after — it rose by 0.031. Neighbour-to-neighbour the worst pair went 0.862 → 0.852. (The earlier render, with segment 1 left raw, scores 0.662 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, Sadness moved +0.553 in the original and +0.535 after conversion — 97 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.94 → 3.26 (+0.32) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.57, 1.15. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.15.
On the corpus-wide percentile scale those become 0.43, 0.93, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 0.027 before conversion and 0.670 after — it rose by 0.643. Neighbour-to-neighbour the worst pair went 0.027 → 0.670. (The earlier render, with segment 1 left raw, scores 0.531 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, Sadness moved +0.554 in the original and +0.536 after conversion — 97 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.81 → 3.03 (+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 Sadness.
The raw scorer output across the chain is -0.00, 0.52, 1.09. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.09.
On the corpus-wide percentile scale those become 0.43, 0.93, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 0.309 before conversion and 0.688 after — it rose by 0.380. Neighbour-to-neighbour the worst pair went 0.373 → 0.689. (The earlier render, with segment 1 left raw, scores 0.598 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.
The emotional move did not survive. Re-scored end to end, Sadness moved +0.552 in the original and -0.472 after conversion — it changed direction. On this chain the corrected audio is not an improvement.
Quality. Mean predicted overall quality across the segments went 2.93 → 3.18 (+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 Sadness.
The raw scorer output across the chain is -0.00, 0.56, 1.10. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.10.
On the corpus-wide percentile scale those become 0.43, 0.93, 0.98 — a total move of +0.55.
That is why the strict rule cannot build this chain. Roughly 90 % of the corpus scores exactly zero on Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.830 before conversion and 0.857 after — it rose by 0.027. Neighbour-to-neighbour the worst pair went 0.830 → 0.847. (The earlier render, with segment 1 left raw, scores 0.696 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, Sadness moved +0.552 in the original and +0.062 after conversion — 11 % of the delta retained, so a meaningful part of the trajectory was flattened.
Quality. Mean predicted overall quality across the segments went 2.61 → 3.09 (+0.48) 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 Sadness.
The raw scorer output across the chain is -0.00, 0.79, 1.39. In the first clip the scorer found no Sadness whatsoever (-0.00); by the last it is at 1.39.
On the corpus-wide percentile scale those become 0.43, 0.96, 0.99 — 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 Sadness, and tied values all collapse onto one point (about 0.45). The clips that genuinely carry Sadness 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 3 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 100–150 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 0.764 before conversion and 0.696 after — it fell by 0.068. Neighbour-to-neighbour the worst pair went 0.764 → 0.696. (The earlier render, with segment 1 left raw, scores 0.749 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, Sadness moved +0.563 in the original and +0.558 after conversion — 99 % of the delta retained, which is essentially all of it.
Quality. Mean predicted overall quality across the segments went 2.96 → 3.14 (+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.