Dancing King Nasir · skill pública Open Skill

Dancing King Nasir

Dancing King Nasir é uma Open Skill pública no Digen. Abra para ver o que este fluxo cria.

@idagsgj

TryOn

1 1 2

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4323009/2026-10-20/qvdvu33b.jpg

---

allowed-tools:
  - generate_music
  - generate_image
  - generate_video
  - extract_audio
  - probe_media
  - mix_clips
  - caption_video
---

# Dancing King Nasir

A two-stage "song + music video" generator: King Nasir drops a bouncy track, then a video of him dancing to that exact track comes out. One prompt in, two assets out (MP3 + MP4).

## When to use
- The user asks for a "Dancing King Nasir" video/song/meme, or wants "King Nasir to dance to X".
- The user wants a short vertical Afrobeat/dancehall music video with King Nasir as the lead.
- The user wants another "episode" of the character in a new style (shmoney, azonto, gqom, amapiano…).

Do **not** use this for generic dance videos with no King Nasir character, or for real-person likeness requests the user has no rights to.

## Reference files
- `references/character.md` — the fixed look of King Nasir (wardrobe, colors, face) and per-style prompt snippets. **Read it before generating the character image**, so every episode keeps the same character.

## Steps

Use `write_todos` to lay out the 4 stages below, and update the whole list as you progress. Keep one item in_progress at a time.

### 1. Lock the brief
Ask at most one round of questions, only if the user hasn't specified:
- **Style** (default: Afrobeat / dancehall hybrid, uptempo, ~110–125 BPM, heavy 808 bass, log-drum, shaker, brass stab, chant-style vocal hook). Offer: shmoney, azonto, gqom, amapiano, highlife.
- **Length**: default 12s of usable video (the audio-driven video channel caps at ~14s). Longer = generate a longer track and cut a 12s highlight.
- **Vibe keywords** (e.g. "crown, gold, celebration, night club").
- **Format**: portrait 9:16 by default (TikTok/Reels); use `landscape` if the user asks for YouTube.

If the user said "just go", pick the defaults above and proceed without asking.

### 2. Generate the track — `generate_music`
```
model='music'
prompt=<style + vibe + instrumentation + vocal-hook description>
lyrics=<2-4 short chant lines, e.g. "King Nasir, oya gba, oya gba / money dance, money dance">   # omit for a purely instrumental track
duration=30            # generate ~30s so there is room to cut a clean 12s hook
output_format='mp3'
name='King Nasir — <style> Track'
entity_type='music'
```
- A chant-style hook is what makes the clip work as a meme; default to one with the character's name in it.
- If the model returns instrumental only, that is fine — do not fake a vocal.

### 3. Cut a clean 12s driver — `probe_media` + `extract_audio`
```
probe_media(url=<music asset_id>)                      # confirm real duration
extract_audio(url=<music asset_id>, segments=[{start:0, duration:12}], format='m4a',
              name='King Nasir — 12s Hook', entity_type='extracted_audios')
```
- Pick the 12s window that lands on the beat drop: start at 0 if unsure, otherwise 4–8s into the track.
- If the track is shorter than 12s, use the whole track and set the video duration to the track's real length.

### 4. Make the character image — `generate_image`
```
model='t2i'
prompt=<character description from references/character.md + the chosen style's outfit and scene>
orientation='portrait'
resolution='1080P'
name='King Nasir — <style> Portrait'
entity_type='characters'
```
- If the user uploaded a photo of the character earlier in the conversation, use `model='i2i'` with that image as `input_urls[0]` instead, and describe the same scene/outfit — this keeps the likeness.
- If the user has a bound preset asset for King Nasir, pass its `asset_id` in `input_urls` and use `model='i2i'`.

### 5. Dance video — `generate_video` (audio-driven)
```
model='ai2v'
input_urls=[<character image asset_id>]
audio_url=<12s audio asset_id>
duration=12
orientation='portrait'
prompt=<dance description: which moves, camera framing, lighting, energy>
name='King Nasir Dancing — <style>'
entity_type='final_videos'
```
- `ai2v` needs **both** an image and an audio track — the dance is driven by the beat, so the music and