Skill

voice-to-lip-sync

Generate a voice take (ElevenLabs or similar) and lip-sync it to an existing video clip of a character speaking. Use for talking-head shots where the visual is already generated.

Install

npx claudepluginhub danielrosehill/claude-code-plugins --plugin ai-video-producer

Tool Access

This skill uses the workspace's default tool permissions.

Preview

Generates a voice line and conforms an existing silent character clip to it.

SKILL.md

Similar Skills

cache-components

Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.

cache-components

139.2k

mcp-builder

9 files

Guides building MCP servers enabling LLMs to interact with external services via tools. Covers best practices, TypeScript/Node (MCP SDK), Python (FastMCP).

anthropics-skills-13

124.2k

canvas-design

20 files

Generates original PNG/PDF visual art via design philosophy manifestos for posters, graphics, and static designs on user request.

anthropics-skills-13

124.2k

Stats

Stars0

Forks0

Last CommitApr 26, 2026

Actions

View Source View Plugin View on GitHub View README

Voice → Lip-Sync Pipeline

Generates a voice line and conforms an existing silent character clip to it.

When to use

Talking-head or dialogue shots where you have a generated character clip but no audio.

Voice-cloning a specific speaker for narration with on-screen presence.

Inputs

The line to be spoken (from scripts/final/script.md or the shot brief).

A character clip in clips/raw/ or generation/image-to-video/.

Voice selection from brief/tools-and-models.md. If a voice clone is in use, source audio lives at assets/voice-clones/<name>.wav.

Steps

Confirm the line. Read it back to the user with character + tone.

Generate voice take. Call the configured voice model (ElevenLabs, etc.). For voice clones, pass the clone source from assets/voice-clones/. Save to generation/voice/NN-shortname-vN.wav. Save script + model + voice ID to generation/prompts/NN-shortname-vN-voice.md. Log it.

Iterate. Play it back (or describe what was generated). Offer to retry with adjusted prosody/emotion settings.

Lip-sync. Pass the approved voice + character clip into the configured lip-sync model. Save to generation/lip-sync/NN-shortname-vN.mp4. Log it.

Verify sync quality. Surface the result. Lip-sync models often need iteration on which face region to lock to or whether the source clip's mouth is occluded.

Accept. Copy to clips/raw/ and suggest /promote-take.

Common gotchas

Source clip face visibility. Lip-sync needs a clear, mostly-frontal mouth. If the character clip cuts away or shows them in profile, the sync will be poor — prefer regenerating the visual with a more frontal angle.

Voice take duration vs clip duration. Voice takes usually need to be slightly shorter than the clip to leave room for natural pauses. If the line is too long, either shorten the script or regenerate the visual longer.

Audio sample rate. Lip-sync models often expect 16kHz or 24kHz mono. Convert with ffmpeg if the voice model outputs 44.1kHz stereo.

Voice → Lip-Sync Pipeline

Generates a voice line and conforms an existing silent character clip to it.

When to use

Talking-head or dialogue shots where you have a generated character clip but no audio.
Voice-cloning a specific speaker for narration with on-screen presence.

Inputs

The line to be spoken (from scripts/final/script.md or the shot brief).
A character clip in clips/raw/ or generation/image-to-video/.
Voice selection from brief/tools-and-models.md. If a voice clone is in use, source audio lives at assets/voice-clones/<name>.wav.

Steps

Confirm the line. Read it back to the user with character + tone.
Generate voice take. Call the configured voice model (ElevenLabs, etc.). For voice clones, pass the clone source from assets/voice-clones/. Save to generation/voice/NN-shortname-vN.wav. Save script + model + voice ID to generation/prompts/NN-shortname-vN-voice.md. Log it.
Iterate. Play it back (or describe what was generated). Offer to retry with adjusted prosody/emotion settings.
Lip-sync. Pass the approved voice + character clip into the configured lip-sync model. Save to generation/lip-sync/NN-shortname-vN.mp4. Log it.
Verify sync quality. Surface the result. Lip-sync models often need iteration on which face region to lock to or whether the source clip's mouth is occluded.
Accept. Copy to clips/raw/ and suggest /promote-take.

Common gotchas

Source clip face visibility. Lip-sync needs a clear, mostly-frontal mouth. If the character clip cuts away or shows them in profile, the sync will be poor — prefer regenerating the visual with a more frontal angle.
Voice take duration vs clip duration. Voice takes usually need to be slightly shorter than the clip to leave room for natural pauses. If the line is too long, either shorten the script or regenerate the visual longer.
Audio sample rate. Lip-sync models often expect 16kHz or 24kHz mono. Convert with ffmpeg if the voice model outputs 44.1kHz stereo.