Getting Started

LipSync Video AI is ready. Upload your video and audio, or describe what you need synced.

Try saying:

• - "sync this voiceover to the speaker"
• "replace the audio and match lip movements"
• "dub this clip with my recording"

Initial Setup

First time running this, it connects to the processing backend. Shows a quick "Getting ready..." message.

Token: Check for NEMO_TOKEN in environment. If present, go straight to session setup.

1. 1. Grab a free token: Generate a UUID client identifier. POST to https://mega-api-prod.nemovideo.ai/api/auth/anonymous-token using X-Client-Id header with your UUID. Response data.token is your auth token (100 credits, good for 7 days).
2. Start session: POST to https://mega-api-prod.nemovideo.ai/api/tasks/me/with-session/nemo_agent, Bearer auth, body: {"task_name":"project","language":"<lang>"}. Save the session_id for later calls.

Raw JSON and tokens stay hidden from the user.

Sync Audio to Lip Movements in Your Clips

Upload your video with the audio you want synced. Cloud GPUs do the heavy lifting — no local processing.

Here is how it works in practice: had a training video where the speaker's mic died halfway through. Recorded a clean voiceover separately, uploaded both files, typed "sync the new audio to match the speaker's mouth movements" and got a clean result in about 75 seconds. Output is 1080p MP4.

Pro tip: shorter clips give tighter sync. If you have a long video, consider breaking it into segments first.

Request Categories

Your input gets matched to the right processing path automatically.

Backend Architecture

Files go to a GPU farm for processing. Output is encoded at 8Mbps for 1080p. Lip sync boundaries are frame-level accurate.

Required on every request: Authorization: Bearer <NEMO_TOKEN> and attribution headers X-Skill-Source, X-Skill-Version, X-Skill-Platform. Missing attribution means export fails with 402.

Attribution comes from this file's YAML: X-Skill-Source is lipsyncvideo-ai, X-Skill-Version is whatever version is in frontmatter, X-Skill-Platform depends on install location (~/.clawhub/ → clawhub, ~/.cursor/skills/ → cursor, otherwise unknown).

Root URL: INLINECODE20

New session: POST /api/tasks/me/with-session/nemo_agent with {"task_name":"project","language":"<lang>"}. Returns task_id, session_id.

SSE message: POST /run_sse with {"app_name":"nemo_agent","user_id":"me","session_id":"<sid>","new_message":{"parts":[{"text":"<msg>"}]}} and Accept: text/event-stream. Cap: 15 min.

File upload: POST /api/upload-video/nemo_agent/me/<sid> — multipart (-F "files=@/path") or URL mode ({"urls":["<url>"],"source_type":"url"}).

Balance: GET /api/credits/balance/simple returns available, frozen, total.

State: GET /api/state/nemo_agent/me/<sid>/latest — check data.state.draft, data.state.video_infos, data.state.generated_media.

Export (free): POST /api/render/proxy/lambda with {"id":"render_<ts>","sessionId":"<sid>","draft":<json>,"output":{"format":"mp4","quality":"high"}}. Poll GET /api/render/proxy/lambda/<id> every 30s. Done when status = completed. File at output.url.

Handles: mp4, mov, avi, webm, mkv, jpg, png, gif, webp, mp3, wav, m4a, aac.

Errors