Communitygithub.com

lfs-video-face-swap

Estimate credits, prepare and start an approved face swap for an uploaded MP4, or check and download an existing video face swap using the Live Face Swap MCP. Use when the user wants Live Face Swap to process a prerecorded video or asks about a Live Face Swap task.

What is lfs-video-face-swap?

lfs-video-face-swap is a Claude Code agent skill that estimate credits, prepare and start an approved face swap for an uploaded MP4, or check and download an existing video face swap using the Live Face Swap MCP. Use when the user wants Live Face Swap to process a prerecorded video or asks about a Live Face Swap task.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/aloow/lfs-video-face-swap/tree/main

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Live Face Swap for uploaded video

Use the connected Live Face Swap MCP tools. If they are unavailable, point the user to the installation instructions or the Live Face Swap MCP guide. Use the connection the user selected. For local testing, keep every call on the local connection.

Estimate and prepare

  • For a price estimate, call quote_video_face_swap with durationSeconds between 1 and 30. Ask for the duration if it is unknown. Report estimatedCredits as an estimate. A quote-only request ends here.
  • When the user wants to prepare a job, call prepare_video_face_swap with no arguments. Keep the returned draftId and share preparationUrl privately. The link expires after 30 minutes. Ask the user to open it with the same Live Face Swap account, select a 1–30 second MP4 with one clearly visible face and a portrait they have permission to use, and approve the exact quote shown on the site. Keep source files in that browser workflow.

Start an approved job

  • After the user confirms that they approved the exact quote on the site and explicitly asks to start, call create_video_face_swap with the same draftId. Starting reserves credits. Save the returned taskId for status checks.
  • If the server reports VIDEO_MCP_APPROVAL_REQUIRED, send the user back to the preparation link. If the draft expired or the quote changed, offer a fresh preparation link and repeat the approval step. If the start request times out or the outcome is unclear, retry with the same draftId to avoid a duplicate job.

Check and download

  • Call get_video_face_swap_result with the existing taskId. For a status-only request, report the current state. While waiting for completion, honor retryAfterSeconds before checking again. If a call hits a rate limit, use the server's retry delay.
  • On completed, report billedCredits when returned and share the private resultUrl with the user. The result page finishes the video in the browser and offers MP4 preview and download. On failed, report the failure and any billed credits returned. A new attempt requires a new approval.

Related Skills