Dance & Celebrate Only (No Singing / No Lip Sync) — Open Skill

Dance & Celebrate Only (No Singing / No Lip Sync)

Dance & Celebrate Only (No Singing / No Lip Sync) is a public Open Skill on Digen. Open it to see what this workflow creates.

@Ragazul

Music

20 0 8

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1856918/2026-09-04/eq6wkgj7.png

# Dance & Celebrate Only – No Singing / No Lip Sync

## When to use
- The user wants a music video, MV scene, dance clip, or party/celebration scene where the visible person/characters only dance, move, celebrate, and react naturally to the music.
- The user uploads a song (with or without vocals) and wants visuals that never sing, rap, speak, mouth lyrics, or lip-sync.
- The user wants to guarantee that no visible person's mouth moves in sync with the lyrics, even when the track has clear vocals.

## Core rule
The music controls body movement, dancing rhythm, energy, gestures, celebration, facial expressions, and pacing. The lyrics must never control mouth movement. Even if the uploaded song contains vocals, rap, singing, or spoken words, the person in the video must not move their lips in synchronization with the audio.

Priority order (highest first):
1. NO SINGING OR LIP SYNC
2. NATURAL DANCING
3. BEAT SYNCHRONIZATION
4. CELEBRATION
5. NATURAL FACIAL EXPRESSIONS
6. CINEMATIC QUALITY

If realistic facial animation conflicts with the no-lip-sync rule, always prioritize NO LIP SYNC.

## Steps

### 1. Gather inputs
- Uploaded audio (the song) — required for audio-driven generation.
- Character/scene reference images (optional) — used as identity/style references.
- User's choices: number of people, scene/setting, dance/celebration style, orientation (landscape 16:9 / portrait 9:16), clip duration, number of scenes.

### 2. Analyze the audio — music only, never vocals
Analyze the uploaded audio ONLY for musical properties:
- rhythm, tempo, beat, bass, drops, energy, musical transitions, mood
Do NOT transcribe lyrics, do NOT use vocal timing, do NOT call transcribe_audio to obtain lyrics, and do NOT map words to mouth movement.
Map: Beat → movement; Bass → body rhythm; Tempo → movement speed; Drops → stronger dance reaction; Breaks → reduced or slower movement; Build-ups → increasing excitement.
Never map: Lyrics → mouth movement; Vocals → singing performance; Rap → rapping character.

### 3. Plan the scene
- Decide characters (1 or more), setting, dance/celebration style, and an energy curve that matches the track.
- If several people appear: none of them may sing, rap, speak, or lip-sync. All visible characters may only dance, celebrate, react, interact naturally, and enjoy the music. Do not assign the vocals to any visible character.

### 4. Generate stills (if needed)
- If no character reference exists, generate a still with generate_image (model='t2i' for text-to-image, or model='i2i' with reference images): a dancer/celebrant in the scene, mouth closed or naturally relaxed, no microphone, no singing pose.
- name="scene_still", entity_type="storyboard_images".

### 5. Generate the video (audio-driven, no lip sync)
- With a character/scene reference image: generate_video with model='ref2v', input_urls=[reference image(s)], audio_url=<uploaded song>, audio_start_time=<start offset if needed>, orientation, duration.
- With a generated still and no reference: use the same ref2v approach with the still, or model='ai2v' (image + audio) if ref2v is not suitable.
- Always pass the audio via audio_url — never stuff it into input_urls.
- Every video prompt must include the strict no-lip-sync direction (see Prompt template below).
- name="dance_clip", entity_type="storyboard_videos".

### 6. Verify
- After generation, check the clip with recognize_video: confirm no lip sync, no mouthing, no singing/rapping poses, mouth closed or naturally relaxed, and dancing/celebration matching the beat.
- If any lip sync or mouthing appears, regenerate with stronger no-lip-sync wording (e.g. "mouth closed, lips still, no singing, no lip sync, no mouthing words").

### 7. Stitch (optional)
- For multi-scene music videos, generate each scene clip, then mix_clips to concatenate them and lay the original song on the cut (audio track with the song URL).
- name="final_music_video", entity_type="final_videos".

## Prompt template (no-lip-sync direction)
I