3D Cinematic Singer MV · öffentlicher Open Skill
3D Cinematic Singer MV
Create a 3D Cinematic music 🎵🎶 video and vibe your day
Music
3 0 12
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4070602/2026-09-27/kbxhmggp.jpg
# 3D Cinematic Singer Music Video Turn a song/idea into a short, vertical, crowd-engaging performance MV: a single 3D-cinematic look kept consistent across every frame. "Believer"-style public performance is the classic case; the same workflow works for any song or setting. ## When to use - User wants a 3D cinematic video of a person singing in public with fans/crowd engaging, to a given song (e.g. "Believer") or a loose idea. - User wants to restyle the singer (look, outfit, stage), change the public setting, or change the song/energy. - User wants to retake/iterate on the last generated shot (re-render, tweak, or a follow-up "same performer, next hook" style edit). ## Steps For multi-step work, first call `write_todos` to list the stages and update the list as you progress. 1. **Clarify (if needed).** Confirm the song title, the singer's look (gender, outfit, vibe), and the public setting (street corner, plaza, rooftop, subway, festival stage…). If the user named no specific song/character or setting, default to a confident lead performer on a music-festival main stage at night, facing a vast, energized crowd, and treat "Believer" as the default song theme. Keep the requested 3D-cinematic style. 2. **Create the keyframe image** with `generate_image`: - `model='t2i'`, `orientation='portrait'` (vertical, short-form), `resolution='1080P'`. - Prompt (English): a single charismatic lead singer mid-performance on a towering music-festival main stage at night, dynamic dramatic pose, a huge sea of fans reaching out/cheering, cinematic 3D render look (high detail, festival light rigs, haze and volumetric lights, shallow depth of field, anamorphic framing, the crowd mass blurring in the foreground for depth). Describe the specific singer look, outfit, setting, and song energy. - `name='<performer> keyframe'`, `entity_type='storyboard_images'`. 3. **Make the soundtrack** with `generate_music`: - `model='music'`, `duration` ~15s (short-form cadence). - Prompt (English): a cinematic, driving pop-rock anthem matching the song's title/theme and the arena-performance energy (e.g. for "Believer": soaring, anthemic, percussion-forward). If the user has real lyrics, pass them via `lyrics`. - `name='<song> soundtrack'`, `entity_type='music'`. 4. **Drive the singing scene** with `generate_video`: - `model='ref2v'` (reference-to-video, audio-driven). `input_urls=[<keyframe asset_id>]` and `audio_url=<music url>` (optionally `audio_start_time`) so the performer appears to sing in sync. - `orientation='portrait'`, `duration` ~10–15s. - Prompt (English): the same performer on the festival main stage, shown mid-performance, clearly singing/mouthing to the track, the vast crowd responding/cheering and moving with the beat, camera subtly pushing in, consistent 3D cinematic style, no text overlays. - `name='<song> MV shot'`, `entity_type='final_videos'`. 5. **Optional: assemble a short MV** with `mix_clips` if the user wants more than one shot: - Pass multiple generated clips as `videos` tracks (concatenated in order), `orientation='portrait'`, `name='<song> MV final'`, `entity_type='final_videos'`. 6. **Reply & iterate.** Summarize the result naturally (don't paste raw URLs). Offer next steps via `set_guidance` (restyle singer, change the setting, make the next hook/shot, adjust the song/energy, add the actual lyrics). ## Notes - Keep ONE consistent 3D-cinematic visual style across images and video; reuse the same keyframe as the ref for any follow-up shot so the performer's identity stays coherent. - Video generation does not support NSFW/adult content on any channel — decline such requests politely. - Report generation failures honestly; never fabricate URLs. If the user dislikes a shot, re-run step 2/4 with the requested tweak. - Intermediate per-step products (keyframe, soundtrack, individual shots) will show on the canvas as normal assets; that's expected.