The production path
A regeneration request identifies a stored section and creates a replacement through the same selected voice workflow. After the provider returns audio, the application validates the replacement, re-merges the complete take, rebuilds the SRT, and promotes the new revision only after the publish checks pass.
- The original published take stays available if generation, merge, storage, or validation fails.
- Revision keys and job guards reduce duplicate or concurrent promotions.
- The regenerated section is highlighted in the rebuilt SRT proof so the change is inspectable.
What it does not promise
The current capability is section-level, not arbitrary sentence-level replacement. A section can contain more than one sentence, and the rebuilt timing can change for the surrounding cues. This boundary is intentional: it keeps the workflow explainable and safer to roll back.
When to use it
Use regeneration for factual corrections, revised openings, changed calls to action, or a pronunciation fix. Listen to the transition into and out of the replacement, then download both artifacts from the new revision.