Quality control

Timing QA

Every translated segment is measured against its source slot. Drift outside the band triggers padding, a bounded tempo fit, or a blocked retry that waits for approved wording.

Drift bands

≤ 8% fit8–15% pad or review> 15% retry

Correction archive

Measured red → approved rewrite → verified green

B2 verified archive

One German attempt overran a 3.800s source slot. Toluva measured it, blocked the next billable call, and resumed only after a hash-bound human-approved shorter revision appeared in B2. Both attempts, both manifests, and the parent/child run relationship remain inspectable.

Attempt 1red

Willkommen bei Toluva. Mit unserer Plattform wird eine einzige Videobotschaft automatisch in vielen verschiedenen Sprachen verfügbar.

Generated: 8.127s
Slot: 3.800s
Drift: +113.87%

Run a9cc6c70…50c9a · manifest and stored bytes verified

Attempt 2green

Willkommen bei Toluva. Eine Botschaft, viele Sprachen.

Generated: 3.576s
Slot: 3.800s
Drift: −5.90%

Run 3e5cdd07…3f1c3 · manifest and stored bytes verified

Genblaze parent/child lineagea9cc6c70…50c9a → 3e5cdd07…3f1c3

Protected term Toluva preserved

German timing report

3 measured segments

3 TTS calls · 189 chars

segment-001−8.88%

One video can reach every team.

Ein Video kann jedes Team erreichen.

Slot: 2.580s
Generated: 2.351s
Final: 2.580s

segment-002+4.49%

Toluva keeps each synthetic voice authorized.

Toluva hält jede synthetische Stimme autorisiert.

Slot: 3.150s
Generated: 3.291s
Final: 3.150s

segment-003−12.52%

When translated speech runs long, Toluva measures the drift and pauses for approved wording.

Wenn übersetzte Sprache lange dauert, misst Toluva die Drift und pausiert für genehmigte Formulierungen.

Slot: 6.689s
Generated: 5.851s
Final: 6.689s

Segment	Source / final translation	Slot	Generated	Final	Drift
segment-00100:00.000 – 00:02.580	One video can reach every team.Ein Video kann jedes Team erreichen.	2.580s	2.351s	2.580s	−8.88%
segment-00200:02.580 – 00:05.730	Toluva keeps each synthetic voice authorized.Toluva hält jede synthetische Stimme autorisiert.	3.150s	3.291s	3.150s	+4.49%
segment-00300:05.730 – 00:12.419	When translated speech runs long, Toluva measures the drift and pauses for approved wording.Wenn übersetzte Sprache lange dauert, misst Toluva die Drift und pausiert für genehmigte Formulierungen.	6.689s	5.851s	6.689s	−12.52%

Source timing preserved with measured correction

Segments 1 and 3 kept their natural delivery and received silence padding. Segment 2 ran 4.49% long, so the audio fan-in applied a bounded 1.0449× tempo fit. No extra TTS call was needed, and every segment still lands on its original boundary.