sc-AB2-k4 — voice-corrected

What was done to this chain. All 4 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 0.913 before conversion and 0.892 after — it fell by 0.022. Neighbour-to-neighbour the worst pair went 0.916 → 0.866. (The earlier render, with segment 1 left raw, scores 0.635 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, Fear moved +0.262 in the original and +0.440 after conversion — 168 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Longing, -0.396 became -0.402.

Quality. Mean predicted overall quality across the segments went 2.97 → 3.06 (+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.

What was done to this chain. All 4 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 0.787 before conversion and 0.709 after — it fell by 0.078. Neighbour-to-neighbour the worst pair went 0.787 → 0.725. (The earlier render, with segment 1 left raw, scores 0.579 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 +0.382 in the original and +0.049 after conversion — 13 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Emotional Numbness, -0.382 became -0.306.

Quality. Mean predicted overall quality across the segments went 2.95 → 3.08 (+0.13) 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.

What was done to this chain. All 4 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 0.869 before conversion and 0.818 after — it fell by 0.051. Neighbour-to-neighbour the worst pair went 0.852 → 0.770. (The earlier render, with segment 1 left raw, scores 0.713 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 +0.398 in the original and +0.339 after conversion — 85 % of the delta retained, which is most of it. On the other named axis, Emotional Numbness, -0.417 became -0.865.

Quality. Mean predicted overall quality across the segments went 2.94 → 3.00 (+0.06) 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.

What was done to this chain. All 4 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 0.876 before conversion and 0.856 after — it fell by 0.020. Neighbour-to-neighbour the worst pair went 0.866 → 0.874. (The earlier render, with segment 1 left raw, scores 0.767 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, Doubt moved +0.263 in the original and +0.000 after conversion — 0 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Infatuation, -0.322 became +0.014.

Quality. Mean predicted overall quality across the segments went 2.78 → 2.95 (+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.

What was done to this chain. All 4 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 0.593 before conversion and 0.618 after — it rose by 0.025. Neighbour-to-neighbour the worst pair went 0.724 → 0.707. (The earlier render, with segment 1 left raw, scores 0.500 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, Triumph moved +0.424 in the original and +0.472 after conversion — 111 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Impatience and Irritability, -0.339 became -0.242.

Quality. Mean predicted overall quality across the segments went 2.94 → 3.13 (+0.19) on the Empathic-Insight head. The corrected render is 48 kHz because SIDON outputs 48 kHz; the original is the 24 kHz source. Some of what you hear as “cleaner” is that bandwidth, not the conversion.

What was done to this chain. All 4 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 0.684 before conversion and 0.670 after — it fell by 0.014. Neighbour-to-neighbour the worst pair went 0.684 → 0.670. (The earlier render, with segment 1 left raw, scores 0.598 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, Emotional Numbness moved +0.261 in the original and +0.477 after conversion — 183 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Concentration, -0.303 became -0.263.

Quality. Mean predicted overall quality across the segments went 2.87 → 3.03 (+0.16) 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.

What was done to this chain. All 4 segments were re-synthesised with ChatterboxVC onto segment 1's voice — including segment 1 itself, converted with itself as the target — and then restored with SIDON. Converting the anchor too is what keeps the room and the reverb the same across the whole chain; leaving it raw put a change of acoustic at the first join. The words, timing and delivery still come from each original clip. The joins are 150–150 ms equal-power crossfades. The chain is normalised as one signal, so the loudness differences between segments are the ones the conversion produced, not a per-clip reset.

Did it unify the voice? On the 250-dimensional Speaker-wavLM-id verification embedding, the worst similarity between any segment and segment 1 was 0.851 before conversion and 0.793 after — it fell by 0.057. Neighbour-to-neighbour the worst pair went 0.715 → 0.669. (The earlier render, with segment 1 left raw, scores 0.711 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, Contemplation moved +0.275 in the original and +0.277 after conversion — 101 % of the delta retained, which is essentially all of it. On the other named axis, Emotional Numbness, -0.258 became -0.184.

Quality. Mean predicted overall quality across the segments went 2.89 → 3.05 (+0.16) 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.

What was done to this chain. All 4 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 0.743 after — it fell by 0.022. Neighbour-to-neighbour the worst pair went 0.874 → 0.857. (The earlier render, with segment 1 left raw, scores 0.688 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 +0.633 in the original and +0.570 after conversion — 90 % of the delta retained, which is essentially all of it. On the other named axis, Intoxication Altered States of Consciousness, -0.286 became -0.397.

Quality. Mean predicted overall quality across the segments went 3.02 → 3.12 (+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.

What was done to this chain. All 4 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 0.796 before conversion and 0.777 after — it fell by 0.018. Neighbour-to-neighbour the worst pair went 0.886 → 0.875. (The earlier render, with segment 1 left raw, scores 0.578 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, Concentration moved +0.361 in the original and +0.312 after conversion — 86 % of the delta retained, which is most of it. On the other named axis, Thankfulness Gratitude, -0.355 became -0.265.

Quality. Mean predicted overall quality across the segments went 2.45 → 3.02 (+0.56) 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.

What was done to this chain. All 4 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 0.885 before conversion and 0.834 after — it fell by 0.050. Neighbour-to-neighbour the worst pair went 0.902 → 0.834. (The earlier render, with segment 1 left raw, scores 0.680 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, Contemplation moved +0.289 in the original and +0.268 after conversion — 93 % of the delta retained, which is essentially all of it. On the other named axis, Jealousy and Envy, -0.397 became -0.339.

Quality. Mean predicted overall quality across the segments went 2.88 → 3.13 (+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.

What was done to this chain. All 4 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 0.729 before conversion and 0.716 after — it fell by 0.014. Neighbour-to-neighbour the worst pair went 0.719 → 0.716. (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.

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 +0.328 in the original and +0.253 after conversion — 77 % of the delta retained, which is most of it. On the other named axis, Longing, -0.344 became -0.381.

Quality. Mean predicted overall quality across the segments went 3.07 → 3.13 (+0.06) 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.

What was done to this chain. All 4 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 0.869 before conversion and 0.913 after — it rose by 0.044. Neighbour-to-neighbour the worst pair went 0.869 → 0.913. (The earlier render, with segment 1 left raw, scores 0.818 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, Doubt moved +0.273 in the original and +0.329 after conversion — 121 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Concentration, -0.301 became -0.208.

Quality. Mean predicted overall quality across the segments went 3.09 → 3.23 (+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.

What was done to this chain. All 4 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 0.805 before conversion and 0.883 after — it rose by 0.078. Neighbour-to-neighbour the worst pair went 0.910 → 0.859. (The earlier render, with segment 1 left raw, scores 0.784 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 +0.524 in the original and +0.466 after conversion — 89 % of the delta retained, which is most of it. On the other named axis, Concentration, -0.318 became -0.366.

Quality. Mean predicted overall quality across the segments went 3.00 → 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.

What was done to this chain. All 4 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 0.848 before conversion and 0.857 after — it rose by 0.009. Neighbour-to-neighbour the worst pair went 0.848 → 0.857. (The earlier render, with segment 1 left raw, scores 0.755 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, Contemplation moved +0.257 in the original and +0.098 after conversion — 38 % of the delta retained, so a meaningful part of the trajectory was flattened. On the other named axis, Concentration, -0.271 became -0.448.

Quality. Mean predicted overall quality across the segments went 3.03 → 3.28 (+0.24) 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.

What was done to this chain. All 4 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 0.941 before conversion and 0.852 after — it fell by 0.089. Neighbour-to-neighbour the worst pair went 0.934 → 0.852. (The earlier render, with segment 1 left raw, scores 0.535 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, Emotional Numbness moved +0.285 in the original and +0.313 after conversion — 110 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Pain, -0.369 became -0.369.

Quality. Mean predicted overall quality across the segments went 2.96 → 3.05 (+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.

What was done to this chain. All 4 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 0.764 before conversion and 0.767 after — it rose by 0.002. Neighbour-to-neighbour the worst pair went 0.806 → 0.756. (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, Intoxication Altered States of Consciousness moved +0.289 in the original and +0.349 after conversion — 121 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Concentration, -0.429 became -0.459.

Quality. Mean predicted overall quality across the segments went 2.65 → 2.83 (+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.

What was done to this chain. All 4 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 0.858 before conversion and 0.883 after — it rose by 0.025. Neighbour-to-neighbour the worst pair went 0.858 → 0.883. (The earlier render, with segment 1 left raw, scores 0.826 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, Shame moved +0.378 in the original and +0.230 after conversion — 61 % of the delta retained. On the other named axis, Contemplation, -0.336 became -0.153.

Quality. Mean predicted overall quality across the segments went 2.94 → 3.04 (+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.

What was done to this chain. All 4 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 0.871 before conversion and 0.887 after — it rose by 0.017. Neighbour-to-neighbour the worst pair went 0.889 → 0.916. (The earlier render, with segment 1 left raw, scores 0.819 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 +0.525 in the original and +0.700 after conversion — 133 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Concentration, -0.296 became -0.306.

Quality. Mean predicted overall quality across the segments went 3.01 → 3.24 (+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.

What was done to this chain. All 4 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 0.822 before conversion and 0.819 after — it fell by 0.002. Neighbour-to-neighbour the worst pair went 0.795 → 0.782. (The earlier render, with segment 1 left raw, scores 0.755 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, Impatience and Irritability moved +0.286 in the original and +0.600 after conversion — 210 % of the delta retained, i.e. the move came out slightly larger after conversion than before. On the other named axis, Astonishment Surprise, -0.275 became -0.238.

Quality. Mean predicted overall quality across the segments went 2.89 → 3.10 (+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.

What was done to this chain. All 4 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 0.726 before conversion and 0.690 after — it fell by 0.036. Neighbour-to-neighbour the worst pair went 0.796 → 0.693. (The earlier render, with segment 1 left raw, scores 0.602 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, Contemplation moved +0.370 in the original and +0.319 after conversion — 86 % of the delta retained, which is most of it. On the other named axis, Pride, -0.378 became -0.490.

Quality. Mean predicted overall quality across the segments went 2.88 → 3.08 (+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.