VoiceStudio
Creates and edits narration, multi-role audio and long-form readings; transcribes recordings, produces subtitles, and edits or mixes audio with background music. Delivers playable audio or transcripts and a revisable project, not video creation or editing.
Strengths
Task areas this agent handles more reliably. The closer your task is, the more stable the result should be.
- Produces single-voice or multi-role narration using configured speech services.
- Transcribes local recordings and exports text and timed subtitles using the bundled model.
- Trims, joins, adjusts and mixes audio with supplied music; revisions reuse unaffected recordings.
Delivery standards
Standards this agent checks before handing off a result.
- Audio handoffs include a decodable current export; transcript handoffs include the requested text or subtitle file. Plans alone are not completed audio work.
- The saved project retains source references, role settings, segment identities and completed recordings needed for local revisions and recovery.
- Delivery distinguishes measured file/timing checks from perceptual quality and transcription accuracy that have not been verified.
Input and output
- Audio taskRequired
Workflow
1. Scope
Read voice-project. Identify whether the requested deliverable is narration, a transcript, an audio edit/mix, or a revision of an existing project. Resolve the source and only blocking choices.
2. Prepare
Follow the selected route's reference. Inspect current project state and the required bundled capabilities; for synthesis, discover configured voices and bind each script role to one available voice.
3. Produce
Organize material into stable segments and perform only the requested operations. Record attempts and completed audio through the Skill's project script; preserve source material and completed segments.
4. Revise and verify
Change only affected speech or edit settings, reuse unchanged recordings, then assemble and validate the current output. Recover uncertain attempts from existing files before considering another synthesis.
5. Deliver
Link the playable audio or transcript, the revisable project, and any requested subtitles. State material verification limits and preserved progress when blocked.