Cinematic Dance MV Producer · публичный навык Open Skill
Cinematic Dance MV Producer
Create a high-energy cinematic English dance music video featuring an adult female pop performer and professional adult backup dancers. Style: modern international pop music video, glamorous, confident, stylish, energetic, premium cinematic production. Scene: a luxurious neon-lit rooftop at night, glowing city skyline in the background, colorful LED lights, light fog, reflective floor. Performance: the singer performs an original English pop/dance song with powerful vocals and catchy rhythm. She performs confident, stylish choreography with the backup dancers, synchronized movements, dynamic formations, smooth transitions and expressive performance. Camera: cinematic close-ups, full-body dance shots, smooth tracking shots, slow-motion moments, dramatic camera movements, occasional wide aerial-style shots. Visuals: glossy fashion styling, sparkling stage lights, vibrant neon atmosphere, realistic skin and hair, natural facial expressions, detailed environment, professional concert lighting. Mood: hot, glamorous, energetic, confident and fashionable — sensual through performance and styling, but non-explicit. Video quality: ultra-realistic, cinematic, 4K, high detail, professional music-video production, smooth motion, realistic physics, sharp focus, high-end color grading. Music: original English pop-dance track, powerful female vocals, catchy hook, deep bass, modern electronic beat, 120–125 BPM. Duration: 10 seconds per generated clip. Maintain the same performer, outfit, hairstyle, face and visual style across every scene.
Music
6 1 8
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3821453/2026-09-29/vm54ahha.png
# Cinematic Dance MV Producer ## When to use - The user wants a high-energy cinematic English dance/pop music video: a female pop performer + professional backup dancers, neon night rooftop, original pop-dance track, 10s clips, consistent performer across scenes. - The user only wants to revise one stage (performer look, song, shots) or continue the next 10s along the same song. ## Steps 1. **Brief (lock the job).** Write one internal sentence of what the user wants this run to do (performer, song, duration, aspect, pictures). Lock defaults for any gap: adult female pop performer, professional backup dancers, neon-lit rooftop at night with glowing city skyline / colorful LED lights / light fog / reflective floor, original English pop-dance song (120–125 BPM, powerful female vocals, catchy hook, deep bass, modern electronic beat), 10s per clip, `output_aspect=portrait` (9:16) unless the user explicitly asks for 16:9/landscape. If the user uploaded a person photo, that photo is the performer — keep face and clothes as photographed (do not invent a new face/wardrobe). If no photo, invent the performer (random female) without asking. If two+ distinct outfits or two+ people and the lead is unclear, ask via `set_guidance` and stop. 2. **Assets (Stage 2).** Generate one image per character and location, never grids, `entity_type` not `storyboard_images`. Use `generate_image` with `model='t2i'` (or `i2i` when editing a reference), `orientation` per `output_aspect`, `resolution='1080P'`. Prompts are English, one natural-language paragraph each: - CHAR (performer): `name='CHAR'`, `entity_type='characters'` — adult female pop performer, glossy fashion styling, confident glamorous look, realistic skin/hair, natural expression. - LOC (rooftop): `name='LOC'`, `entity_type='locations'` — luxurious neon-lit rooftop at night, glowing city skyline, colorful LED lights, light fog, reflective floor, professional concert lighting. - If the user uploaded a performer photo, use that URI for CHAR and skip `generate_image` for that row. 3. **Music (Stage 3).** If the user uploaded usable audio, skip generation, then `transcribe_audio` to lock `audio_duration`. Otherwise call `generate_music` with `model='music'`, `prompt` in English (original English pop-dance, 120–125 BPM, powerful female vocals, catchy hook, deep bass, modern electronic beat), and write the sung lyrics in `lyrics=` (start with `[Verse]`; never leave `lyrics=` empty). `name='song'`, `entity_type='music'`. Empty URI or failed take → one remake, then continue. 4. **Clips (Stage 4).** Generate 10s clips, `orientation` per `output_aspect`, `resolution='1080P'`, keeping the same performer/outfit/hairstyle/face/style across every scene. Two engines: - **Singing/performance on screen** → `ai2v`: first make one hidden still of the locked CHAR in the LOC via `generate_image` (`model='i2i'`, CHAR as `[0]`, `add_to_canvas=false`), then `generate_video(model='ai2v', input_urls=[still], audio_url=<song URI>, audio_start_time=0, duration=10)`. Do not pass the raw CHAR/LOC to the video call; do not show the still. - **Dance/backup, mouth closed** → `ref2v`: `generate_video(model='ref2v', input_urls=[CHAR, LOC], duration=10)`, no audio. - `name='clip-N'`, `entity_type='storyboard_videos'`. Do not write `subtitle`/`karaoke`/`caption`/`on-screen lyric` into any video prompt. 5. **Mix (Stage 5).** Stitch the clips into the final MV with `mix_clips`, clips in order, `name='final MV'`, `entity_type='final_videos'`, `orientation` per `output_aspect`. The song lands on the finished MV. 6. **Reply.** Summarize in natural language (performer, song, clips, final MV). Do not paste large URL blocks. ## Notes - All image/video/music-instruction prompts are English; only sung lyrics in `lyrics=` keep the user's language. - SFW default: everything fully clothed and family-safe unless the user explicitly asks for adult content. Video generation does not support NSFW on any channel. - Never