beach-cafe-model-video · öffentlicher Open Skill
beach-cafe-model-video
beach-cafe-model-video ist ein öffentlicher Open Skill auf Digen. Öffne ihn, um zu sehen, was dieser Workflow erstellt.
ShortDrama
16 1 10
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4167672/2026-10-10/7tzw7kdp.jpg
# Beachside Ice Cream Coffee Break Turn a model photo into a 15-second vertical (9:16) beachside café video with 4 scenes: walking to the kiosk → asking the barista → receiving & paying → enjoying the first taste. The model's face, hairstyle, skin tone, and outfit are locked from the reference image and kept consistent across every scene. ## When to use - The user wants a 15-second beachside ice-cream-coffee café video of their model from a photo. - The user wants to redo, tweak (change a scene, the on-screen text, the music), or continue one of these beach café model videos. ## Reference files - `references/beach-cafe-script.md`: the full 4-scene script with per-scene AI video prompts and on-screen text. Read it before generating. ## Steps 1. Get the model reference image. If the user uploaded one this turn, use it; otherwise ask them to upload a clear photo of the model (face, hair, and outfit visible). Optionally run `recognize_image` on it to note the exact outfit/hairstyle so the prompts stay faithful. 2. Read `references/beach-cafe-script.md` for the 4 scene prompts and on-screen text. 3. Generate the 4 scene clips with `generate_video`, one call per scene, using: - `model='ref2v'` (reference-to-video, preserves the model's identity) - `input_urls=[<model reference image>]` for every scene - `orientation='portrait'` (9:16 vertical) - `prompt` = the scene's AI video prompt from the reference file, always restating "exact same face, hair, and outfit as the reference image" - `name` = e.g. `scene1-walk`, `scene2-order`, `scene3-pay`, `scene4-taste`; `entity_type='storyboard_videos'` 4. Generate a short upbeat, breezy acoustic background track with `generate_music` (prompt: breezy acoustic guitar, warm tropical beach café vibe, ~15s). `name` = `beach-cafe-music`, `entity_type='music'`. 5. Stitch the 4 clips in order with `mix_clips`, one clip per scene, each with its burned-in on-screen text from the reference file (subtitles `items`, `position='bottom'`), and add the music track as the clip audio. `orientation='portrait'`, `name` = `beach-cafe-final`, `entity_type='final_videos'`. 6. Reply with the final video in natural language (no big URL dumps). If a scene's output breaks the model's identity, regenerate that scene and re-mix. ## Notes - The model's face, hair, skin tone, and outfit must stay identical across all 4 scenes — always pass the same reference image and restate the lock in every prompt. - Video generation does not support NSFW/adult content; decline such requests politely. - Report failures honestly and never fabricate URLs.