Time voice-overs with AI
Extract word timings from recorded narration and convert seconds into project frames for animation and motion-design timing.
Time voice-overs with AI when a visual event needs to land on a spoken word. vo-timing explains how to obtain word timestamps from an audio file and convert seconds into project frames.
The guide is aimed at motion-design work rather than general transcription. It covers narration inside an edit, composition offsets and different frame-numbering conventions. You still check the timestamps by listening and verify the result in the actual scene.
What it does
- Extract word-level timestamps from recorded narration
- Convert seconds into frame numbers
- Prepare timing references for animated events
- Extract narration from an edited video
- Account for a composition's starting offset
- Distinguish zero-based frames from Houdini's frame convention
Install and use
- Install the skill in an agent with local execution tools. You need Python and faster-whisper; embedded video audio also needs ffmpeg. Speech models download separately.
- Provide the audio and project frame rate. Ask: “Extract word timestamps and convert them into frames.” The guide's model examples target English speech.
- For an edit, request: “Extract the audio first and preserve the offset between the edit and composition.”
- Listen against the reported timings, check the application's frame origin and keep production-specific results in private project notes.
FAQ
Source
Made by dima-rdnv. Original skill: vo-timing, published on GitHub under the MIT licence.
View on GitHub