mage-lipsync

Installation
SKILL.md

Mage Lipsync

Turn one image and one audio clip into a talking video: the subject's mouth follows the audio, and everything else about the image holds. This is the Lipsync app as a skill: the same prompt and models, run through the Mage connector.

Before you start

  1. Mage tools: get_model, estimate_cost, generate, get_request, list_characters, list_references, create_reference, and create_upload for local files. If they are missing, ask the user to connect Mage (Claude: Customize → Connectors → Add custom connector, https://mcp.mage.space/mcp; Claude Code: /mcp to sign in if the Mage plugin is installed, otherwise claude mcp add --transport http mage https://mcp.mage.space/mcp; others: https://www.mage.space/mcp), then wait.
  2. Inputs are links: https URLs of the files themselves, data URLs under about 3 MB, uploads from disk (create_upload, when you can run commands), or earlier Mage results. Files attached to the chat never reach Mage: ask for a direct link.
  3. Consent. Harmful deepfakes are forbidden on Mage. Only use a real person's face, body, or voice with that person's permission, and never to deceive, harass, or sexualize. When the user hasn't said whose likeness it is, state this rule in one line with the price instead of questioning them; if they say they lack permission, stop.

UX rules

  1. Don't ask for what the request already gives. Ask for everything that is missing in one compact question, then wait.
  2. Send the app's prompt below word for word. Fill in only the marked slots; don't rewrite, translate, or add to it.
  3. Always quote the model and the price in Gems and wait for a yes before generating.
  4. Deliver the link with a one-line label (model, size or length, Gems spent). No prompts, ids, or JSON. Reply in the user's language. Links last 30 days.

Inputs

Installs
51
GitHub Stars
4
First Seen
Oct 8, 2026