B1 at chain length k=3, all corpora, at the mining floor.
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 chain comes from the one-sided rule: only Sexual Lust had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Sexual Lust around average — 0.54, higher than 54 % of clips in this corpus — and ends with it at the very top of the corpus at 0.92, higher than 92 % of clips in this corpus. That is a total rise of 0.39.
Nothing was asked of the other axis, and in fact Pain barely moves at all, sitting near 0.90 throughout.
It takes 3 clips to get there. Clip to clip the moves are +0.16, then +0.23 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.709 before conversion and 0.658 after — it fell by 0.051. Neighbour-to-neighbour the worst pair went 0.755 → 0.697. (The earlier render, with segment 1 left raw, scores 0.538 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, Sexual Lust moved +0.385 in the original and +0.802 after conversion — 208 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Pain, +0.013 became +0.011.
Quality. Mean predicted overall quality across the segments went 2.46 → 2.72 (+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 chain comes from the one-sided rule: only Contempt had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Contempt clearly present — 0.70, higher than 70 % of clips in this corpus — and ends with it at the very top of the corpus at 0.95, higher than 95 % of clips in this corpus. That is a total rise of 0.25.
Nothing was asked of the other axis, and in fact Relief drifts down from 1.00 to 0.53 (-0.46), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.25, then +0.01 — most of the change happening immediately, then levelling off.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.73 against the first clip, where 1.00 would mean an identical voice. That is below the 0.80 threshold the mining used — treat the “same speaker” claim here with caution. Neighbouring clips score at worst 0.73 against each other.
Voice consistency: these clips are separate recordings joined together and the match is loose (0.73, under the 0.80 threshold), so the voice may audibly change 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 0.732 before conversion and 0.764 after — it rose by 0.033. Neighbour-to-neighbour the worst pair went 0.732 → 0.822. (The earlier render, with segment 1 left raw, scores 0.612 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, Contempt moved +0.255 in the original and +0.109 after conversion — 43 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Relief, -0.432 became -0.318.
Quality. Mean predicted overall quality across the segments went 2.56 → 2.96 (+0.41) 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 chain comes from the one-sided rule: only Fatigue Exhaustion had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Fatigue Exhaustion around average — 0.57, higher than 57 % of clips in this corpus — and ends with it strongly present at 0.87, higher than 87 % of clips in this corpus. That is a total rise of 0.30.
Nothing was asked of the other axis, and in fact Contemplation drifts down from 0.75 to 0.48 (-0.27), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.14, then +0.16 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.807 before conversion and 0.764 after — it fell by 0.043. Neighbour-to-neighbour the worst pair went 0.869 → 0.817. (The earlier render, with segment 1 left raw, scores 0.788 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, Fatigue Exhaustion moved +0.303 in the original and +0.118 after conversion — 39 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Contemplation, -0.272 became -0.094.
Quality. Mean predicted overall quality across the segments went 3.02 → 3.11 (+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 chain comes from the one-sided rule: only Infatuation had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Infatuation around average — 0.51, right about the corpus median — and ends with it strongly present at 0.78, higher than 78 % of clips in this corpus. That is a total rise of 0.28.
Nothing was asked of the other axis, and in fact Sexual Lust drifts down from 0.82 to 0.01 (-0.81), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.06, then +0.22 — a slow start, with most of the change arriving in the final step.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? Not directly measured. What does exist is a timbre similarity of 0.91 against the first clip, which describes how alike the voices sound rather than whether they are the same person. It sits on a different scale from the identity check (corpus-wide the timbre numbers run much higher), so it cannot be read against the 0.80 identity threshold and is given here without a pass or fail.
Voice consistency: these clips are separate recordings joined together. Speaker identity was not measured for this sample; the available timbre similarity of 0.91 says the voices sound broadly alike but is not a same-person check. You may notice 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 0.862 before conversion and 0.855 after — it fell by 0.007. Neighbour-to-neighbour the worst pair went 0.901 → 0.859. (The earlier render, with segment 1 left raw, scores 0.763 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, Infatuation moved +0.277 in the original and +0.062 after conversion — 22 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Sexual Lust, -0.809 became -0.441.
Quality. Mean predicted overall quality across the segments went 2.67 → 2.97 (+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 chain comes from the one-sided rule: only Longing had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Longing clearly present — 0.61, higher than 61 % of clips in this corpus — and ends with it at the very top of the corpus at 0.97, higher than 97 % of clips in this corpus. That is a total rise of 0.36.
Nothing was asked of the other axis, and in fact Thankfulness Gratitude drifts down from 0.76 to 0.24 (-0.52), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.11, then +0.24 — a fairly even climb, though some clips carry more of the change than others.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.851 before conversion and 0.840 after — it fell by 0.011. Neighbour-to-neighbour the worst pair went 0.851 → 0.859. (The earlier render, with segment 1 left raw, scores 0.789 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, Longing moved +0.356 in the original and +0.258 after conversion — 73 % of the delta retained, which is most of it. On the other named axis, Thankfulness Gratitude, -0.522 became -0.660.
Quality. Mean predicted overall quality across the segments went 3.17 → 3.27 (+0.10) 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 chain comes from the one-sided rule: only Pain had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Pain around average — 0.54, higher than 54 % of clips in this corpus — and ends with it strongly present at 0.82, higher than 82 % of clips in this corpus. That is a total rise of 0.28.
Nothing was asked of the other axis, and in fact Astonishment Surprise drifts down from 0.89 to 0.52 (-0.37), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.17, then +0.10 — a fairly even climb, though some clips carry more of the change than others.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.818 before conversion and 0.816 after — it fell by 0.002. Neighbour-to-neighbour the worst pair went 0.799 → 0.816. (The earlier render, with segment 1 left raw, scores 0.699 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, Pain moved +0.276 in the original and +0.369 after conversion — 134 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Astonishment Surprise, -0.368 became -0.325.
Quality. Mean predicted overall quality across the segments went 2.89 → 3.01 (+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 chain comes from the one-sided rule: only Pride had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Pride clearly present — 0.69, higher than 69 % of clips in this corpus — and ends with it at the very top of the corpus at 0.98, higher than 98 % of clips in this corpus. That is a total rise of 0.30.
Nothing was asked of the other axis, and in fact Disappointment drifts down from 0.98 to 0.39 (-0.59), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.18, then +0.12 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.920 before conversion and 0.925 after — it rose by 0.005. Neighbour-to-neighbour the worst pair went 0.890 → 0.923. (The earlier render, with segment 1 left raw, scores 0.876 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, Pride moved +0.297 in the original and +0.256 after conversion — 86 % of the delta retained, which is most of it. On the other named axis, Disappointment, -0.589 became -0.595.
Quality. Mean predicted overall quality across the segments went 3.01 → 3.32 (+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 chain comes from the one-sided rule: only Emotional Numbness had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Emotional Numbness around average — 0.56, higher than 56 % of clips in this corpus — and ends with it strongly present at 0.84, higher than 84 % of clips in this corpus. That is a total rise of 0.28.
Nothing was asked of the other axis, and in fact Infatuation drifts down from 0.87 to 0.65 (-0.22), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.10, then +0.17 — a fairly even climb, though some clips carry more of the change than others.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.88 against the first clip, where 1.00 would mean an identical voice. That is above the 0.80 threshold the mining used — very likely one person. Neighbouring clips score at worst 0.92 against each other.
Voice consistency: these clips are separate recordings joined together, matching at 0.88. You may notice the voice shift a little 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 0.863 before conversion and 0.912 after — it rose by 0.049. Neighbour-to-neighbour the worst pair went 0.902 → 0.923. (The earlier render, with segment 1 left raw, scores 0.862 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.
The emotional move did not survive. Re-scored end to end, Emotional Numbness moved +0.291 in the original and -0.194 after conversion — it changed direction. On this chain the corrected audio is not an improvement. On the other named axis, Infatuation, -0.221 became -0.201.
Quality. Mean predicted overall quality across the segments went 3.08 → 3.36 (+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 chain comes from the one-sided rule: only Embarrassment had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Embarrassment strongly present — 0.79, higher than 79 % of clips in this corpus — and ends with it at the very top of the corpus at 1.00, virtually no clip in this corpus scores higher. That is a total rise of 0.20.
Nothing was asked of the other axis, and in fact Fear drifts down from 0.99 to 0.67 (-0.33), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.11, then +0.10 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.19 against the first clip, where 1.00 would mean an identical voice. That is below the 0.80 threshold the mining used — treat the “same speaker” claim here with caution. Neighbouring clips score at worst 0.22 against each other.
Voice consistency: these clips are separate recordings joined together and the match is loose (0.19, under the 0.80 threshold), so the voice may audibly change 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 0.166 before conversion and 0.619 after — it rose by 0.453. Neighbour-to-neighbour the worst pair went 0.211 → 0.710. (The earlier render, with segment 1 left raw, scores 0.581 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, Embarrassment moved +0.201 in the original and +0.137 after conversion — 68 % of the delta retained. On the other named axis, Fear, -0.328 became -0.720.
Quality. Mean predicted overall quality across the segments went 2.87 → 3.06 (+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 chain comes from the one-sided rule: only Anger had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Anger strongly present — 0.76, higher than 76 % of clips in this corpus — and ends with it at the very top of the corpus at 0.99, higher than 99 % of clips in this corpus. That is a total rise of 0.24.
Nothing was asked of the other axis, and in fact Pride drifts down from 0.99 to 0.91 (-0.08), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.03, then +0.20 — a slow start, with most of the change arriving in the final step.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? Not directly measured. What does exist is a timbre similarity of 0.19 against the first clip, which describes how alike the voices sound rather than whether they are the same person. It sits on a different scale from the identity check (corpus-wide the timbre numbers run much higher), so it cannot be read against the 0.80 identity threshold and is given here without a pass or fail.
Voice consistency: these clips are separate recordings joined together. Speaker identity was not measured for this sample; the available timbre similarity of 0.19 says the voices sound broadly alike but is not a same-person check. You may notice 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 0.330 before conversion and 0.480 after — it rose by 0.151. Neighbour-to-neighbour the worst pair went 0.330 → 0.480. (The earlier render, with segment 1 left raw, scores 0.436 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, Anger moved +0.236 in the original and +0.098 after conversion — 42 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Pride, -0.079 became -0.357.
Quality. Mean predicted overall quality across the segments went 2.72 → 3.06 (+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 chain comes from the one-sided rule: only Infatuation had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Infatuation clearly present — 0.60, higher than 60 % of clips in this corpus — and ends with it at the very top of the corpus at 0.93, higher than 93 % of clips in this corpus. That is a total rise of 0.33.
Nothing was asked of the other axis, and in fact Helplessness drifts down from 0.96 to 0.76 (-0.20), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.24, then +0.10 — a fairly even climb, though some clips carry more of the change than others.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.87 against the first clip, where 1.00 would mean an identical voice. That is above the 0.80 threshold the mining used — very likely one person. Neighbouring clips score at worst 0.87 against each other.
Voice consistency: these clips are separate recordings joined together, matching at 0.87. You may notice the voice shift a little 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 0.808 before conversion and 0.739 after — it fell by 0.069. Neighbour-to-neighbour the worst pair went 0.808 → 0.739. (The earlier render, with segment 1 left raw, scores 0.760 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, Infatuation moved +0.362 in the original and +0.147 after conversion — 41 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Helplessness, -0.202 became -0.031.
Quality. Mean predicted overall quality across the segments went 2.91 → 3.13 (+0.21) 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 chain comes from the one-sided rule: only Sourness had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Sourness strongly present — 0.76, higher than 76 % of clips in this corpus — and ends with it at the very top of the corpus at 1.00, virtually no clip in this corpus scores higher. That is a total rise of 0.24.
Nothing was asked of the other axis, and in fact Bitterness climbs from 0.94 to 0.99 (+0.05), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.21, then +0.03 — most of the change happening immediately, then levelling off.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.78 against the first clip, where 1.00 would mean an identical voice. That is below the 0.80 threshold the mining used — treat the “same speaker” claim here with caution. Neighbouring clips score at worst 0.83 against each other.
Voice consistency: these clips are separate recordings joined together and the match is loose (0.78, under the 0.80 threshold), so the voice may audibly change 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 0.677 before conversion and 0.807 after — it rose by 0.130. Neighbour-to-neighbour the worst pair went 0.675 → 0.794. (The earlier render, with segment 1 left raw, scores 0.513 here.) This chain started 0.50-0.70 — audibly different, 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, Sourness moved +0.245 in the original and +0.184 after conversion — 75 % of the delta retained, which is most of it. On the other named axis, Bitterness, +0.058 became +0.183.
Quality. Mean predicted overall quality across the segments went 2.85 → 3.29 (+0.44) 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 chain comes from the one-sided rule: only Infatuation had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Infatuation clearly present — 0.70, higher than 70 % of clips in this corpus — and ends with it at the very top of the corpus at 0.97, higher than 97 % of clips in this corpus. That is a total rise of 0.27.
Nothing was asked of the other axis, and in fact Contemplation barely moves at all, sitting near 0.98 throughout.
It takes 3 clips to get there. Clip to clip the moves are +0.17, then +0.10 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? Not directly measured. What does exist is a timbre similarity of 0.87 against the first clip, which describes how alike the voices sound rather than whether they are the same person. It sits on a different scale from the identity check (corpus-wide the timbre numbers run much higher), so it cannot be read against the 0.80 identity threshold and is given here without a pass or fail.
Voice consistency: these clips are separate recordings joined together. Speaker identity was not measured for this sample; the available timbre similarity of 0.87 says the voices sound broadly alike but is not a same-person check. You may notice 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 0.729 after — it fell by 0.022. Neighbour-to-neighbour the worst pair went 0.810 → 0.776. (The earlier render, with segment 1 left raw, scores 0.694 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, Infatuation moved +0.274 in the original and +0.474 after conversion — 173 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Contemplation, -0.021 became -0.024.
Quality. Mean predicted overall quality across the segments went 2.95 → 3.18 (+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 chain comes from the one-sided rule: only Pride had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Pride clearly present — 0.68, higher than 68 % of clips in this corpus — and ends with it at the very top of the corpus at 0.99, higher than 99 % of clips in this corpus. That is a total rise of 0.31.
Nothing was asked of the other axis, and in fact Anger barely moves at all, sitting near 0.95 throughout.
It takes 3 clips to get there. Clip to clip the moves are +0.17, then +0.14 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.25 against the first clip, where 1.00 would mean an identical voice. That is below the 0.80 threshold the mining used — treat the “same speaker” claim here with caution. Neighbouring clips score at worst 0.25 against each other.
Voice consistency: these clips are separate recordings joined together and the match is loose (0.25, under the 0.80 threshold), so the voice may audibly change 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 0.302 before conversion and 0.840 after — it rose by 0.539. Neighbour-to-neighbour the worst pair went 0.302 → 0.840. (The earlier render, with segment 1 left raw, scores 0.736 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, Pride moved +0.311 in the original and +0.722 after conversion — 232 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Anger, +0.000 became +0.169.
Quality. Mean predicted overall quality across the segments went 2.84 → 3.28 (+0.45) 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 chain comes from the one-sided rule: only Concentration had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Concentration clearly present — 0.73, higher than 73 % of clips in this corpus — and ends with it at the very top of the corpus at 0.99, higher than 99 % of clips in this corpus. That is a total rise of 0.26.
Nothing was asked of the other axis, and in fact Hope Enthusiasm Optimism drifts down from 0.97 to 0.82 (-0.15), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.14, then +0.12 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? Not directly measured. What does exist is a timbre similarity of 0.86 against the first clip, which describes how alike the voices sound rather than whether they are the same person. It sits on a different scale from the identity check (corpus-wide the timbre numbers run much higher), so it cannot be read against the 0.80 identity threshold and is given here without a pass or fail.
Voice consistency: these clips are separate recordings joined together. Speaker identity was not measured for this sample; the available timbre similarity of 0.86 says the voices sound broadly alike but is not a same-person check. You may notice 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 0.860 before conversion and 0.868 after — it rose by 0.008. Neighbour-to-neighbour the worst pair went 0.860 → 0.868. (The earlier render, with segment 1 left raw, scores 0.654 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, Concentration moved +0.261 in the original and +0.268 after conversion — 103 % of the delta retained, which is essentially all of it. On the other named axis, Hope Enthusiasm Optimism, -0.151 became -0.231.
Quality. Mean predicted overall quality across the segments went 2.39 → 3.20 (+0.81) 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 chain comes from the one-sided rule: only Fear had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Fear clearly present — 0.73, higher than 73 % of clips in this corpus — and ends with it at the very top of the corpus at 1.00, virtually no clip in this corpus scores higher. That is a total rise of 0.26.
Nothing was asked of the other axis, and in fact Pride drifts down from 0.77 to 0.24 (-0.53), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.23, then +0.03 — most of the change happening immediately, then levelling off.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.783 before conversion and 0.626 after — it fell by 0.157. Neighbour-to-neighbour the worst pair went 0.819 → 0.679. (The earlier render, with segment 1 left raw, scores 0.620 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, Fear moved +0.264 in the original and +0.139 after conversion — 53 % of the delta retained. On the other named axis, Pride, -0.534 became +0.222.
Quality. Mean predicted overall quality across the segments went 2.37 → 2.58 (+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 chain comes from the one-sided rule: only Hope Enthusiasm Optimism had to get where it was going, by at least 0.20. The other emotion was left completely free.
The chain starts with Hope Enthusiasm Optimism clearly present — 0.73, higher than 73 % of clips in this corpus — and ends with it at the very top of the corpus at 0.97, higher than 97 % of clips in this corpus. That is a total rise of 0.25.
Nothing was asked of the other axis, and in fact Contemplation drifts down from 0.92 to 0.72 (-0.19), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.13, then +0.12 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.92 against the first clip, where 1.00 would mean an identical voice. That is a strong match — almost certainly one person throughout. Neighbouring clips score at worst 0.95 against each other.
Voice consistency: these clips are separate recordings joined together. The measured match is tight (0.92), so any shift should be subtle — but you may still notice the voice change slightly from segment to segment. 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 0.890 before conversion and 0.866 after — it fell by 0.024. Neighbour-to-neighbour the worst pair went 0.931 → 0.861. (The earlier render, with segment 1 left raw, scores 0.648 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, Hope Enthusiasm Optimism moved +0.248 in the original and +0.172 after conversion — 70 % of the delta retained. On the other named axis, Contemplation, -0.197 became -0.087.
Quality. Mean predicted overall quality across the segments went 2.86 → 3.20 (+0.35) 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 chain comes from the one-sided rule: only Astonishment Surprise had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Astonishment Surprise around average — 0.52, higher than 52 % of clips in this corpus — and ends with it at the very top of the corpus at 0.95, higher than 95 % of clips in this corpus. That is a total rise of 0.43.
Nothing was asked of the other axis, and in fact Interest drifts down from 0.88 to 0.57 (-0.31), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.23, then +0.20 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
Same speaker? The least similar clip scores 0.15 against the first clip, where 1.00 would mean an identical voice. That is below the 0.80 threshold the mining used — treat the “same speaker” claim here with caution. Neighbouring clips score at worst 0.15 against each other.
Voice consistency: these clips are separate recordings joined together and the match is loose (0.15, under the 0.80 threshold), so the voice may audibly change 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 0.168 before conversion and 0.658 after — it rose by 0.490. Neighbour-to-neighbour the worst pair went 0.148 → 0.604. (The earlier render, with segment 1 left raw, scores 0.513 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, Astonishment Surprise moved +0.427 in the original and +0.581 after conversion — 136 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Interest, -0.317 became -0.358.
Quality. Mean predicted overall quality across the segments went 2.62 → 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 chain comes from the one-sided rule: only Shame had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Shame around average — 0.50, right about the corpus median — and ends with it at the very top of the corpus at 0.93, higher than 93 % of clips in this corpus. That is a total rise of 0.43.
Nothing was asked of the other axis, and in fact Doubt drifts down from 0.91 to 0.83 (-0.08), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.22, then +0.21 — an even, steady climb — each clip carries about the same share.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.779 before conversion and 0.846 after — it rose by 0.067. Neighbour-to-neighbour the worst pair went 0.693 → 0.834. (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, Shame moved +0.428 in the original and +0.710 after conversion — 166 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Doubt, -0.080 became +0.110.
Quality. Mean predicted overall quality across the segments went 3.08 → 3.17 (+0.10) 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 chain comes from the one-sided rule: only Hope Enthusiasm Optimism had to get where it was going, by at least 0.25. The other emotion was left completely free.
The chain starts with Hope Enthusiasm Optimism clearly present — 0.63, higher than 63 % of clips in this corpus — and ends with it at the very top of the corpus at 0.94, higher than 94 % of clips in this corpus. That is a total rise of 0.31.
Nothing was asked of the other axis, and in fact Astonishment Surprise drifts down from 0.96 to 0.89 (-0.07), which the rule did not require.
It takes 3 clips to get there. Clip to clip the moves are +0.23, then +0.08 — most of the change happening immediately, then levelling off.
No single step is larger than the 0.25 cap, which is exactly what stops this being a jump cut: the change has to be spread across the clips instead of landing all at once.
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 0.888 before conversion and 0.849 after — it fell by 0.039. Neighbour-to-neighbour the worst pair went 0.889 → 0.831. (The earlier render, with segment 1 left raw, scores 0.743 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, Hope Enthusiasm Optimism moved +0.305 in the original and +0.112 after conversion — 37 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Astonishment Surprise, -0.065 became -0.257.
Quality. Mean predicted overall quality across the segments went 3.01 → 3.31 (+0.30) 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.