ViralMiss — Open Skill
ViralMiss
ViralMiss is a public Open Skill on Digen. Open it to see what this workflow creates.
Roleplay
1 0 1
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1747129/2026-10-21/qz7z77wp.jpg
--- allowed-tools: - generate_image - generate_video - generate_music - generate_tts - mix_clips - create_document - update_document - get_document - memory_query - memory_store --- # AI Lifestyle Video Creator ## When to use - The user wants a vertical (9:16) lifestyle video for Digen: AI influencer dancing, morning routine, get-ready-with-me, fashion, daily vlog, or a lifestyle moment. - The user wants a concept with a clear visual hook, smooth motion, a satisfying ending, and ready-to-use prompts that are easy to produce on a phone. - The user wants a consistent AI influencer character across clips. ## Reference files (read on demand, don't read all at once) - `references/lifestyle-formats.md` — playbook for each format (dancing, morning routine, GRWM, fashion, vlog, lifestyle). Read the matching section before writing a concept. - `references/prompt-library.md` — reusable hook, character-sheet, scene-image, and video-motion prompt templates. Read when drafting prompts. - `references/music-guide.md` — vibe-matched music prompts. Read when choosing music. ## Steps 1. **Clarify the brief** (ask only what's missing): which lifestyle format, vibe/tone, length, and any reference idea. If the user gave a full brief, skip to step 2. 2. **Check memory first**: call `memory_query` for the user's preferred formats, character look, vibe, and past projects before proposing anything. 3. **Concept + hook + ending**: propose 1–2 original concepts, each with a clear visual hook, smooth motion plan, and a satisfying ending (see `references/prompt-library.md`). Let the user pick one before generating assets. 4. **Consistent character**: write a short influencer description (name, look, outfit, personality, consistent visual details). Generate a character sheet with `generate_image(model='t2i', prompt=<character sheet prompt>, orientation='portrait', name="<Name> sheet", entity_type="characters")`. 5. **Scene images**: for each scene, generate a key image with `generate_image(model='t2i', prompt=<scene prompt>, orientation='portrait', name="Scene <n>", entity_type="storyboard_images")`. Reuse the character sheet as a reference (`model='i2i'`, `input_urls=[<character sheet asset_id>]`) when the influencer must stay consistent. 6. **Video clips**: generate one clip per scene. For consistent character use `generate_video(model='ref2v', input_urls=[<character sheet>], prompt=<motion description>, orientation='portrait', name="Clip <n>", entity_type="storyboard_videos")`. For a single scene image use `model='i2v'` with that image as the start frame. Keep each clip short (3–8s) and phone-friendly. 7. **Music**: generate a vibe-matched track with `generate_music(prompt=<music prompt from music-guide>, name="Soundtrack", entity_type="music")`. Optionally add a voiceover with `generate_tts(model='tts.design', text=<line>, voice_instruction=<English voice description>, name="Voiceover", entity_type="voiceovers")`. 8. **Assemble**: stitch clips in order with `mix_clips(clips=[...], name="<title> final", entity_type="final_videos")`, adding the music (and optional voiceover) as audio tracks. Match `orientation='portrait'` to the source clips. 9. **Remember**: `memory_store` the user's preferred formats, character look, and vibe so future sessions pick up where they left off. ## Canvas asset conventions Use these fixed `entity_type` values per stage (don't reuse across stages): - Character sheets → `characters` - Scene key images → `storyboard_images` - Generated clips → `storyboard_videos` - Final assembled video → `final_videos` - Music → `music` - Voiceover → `voiceovers` - Concept/script documents → `scripts` ## Memory - **Query** (`memory_query`) at the start of every session for the user's preferred formats, character look, and vibe. - **Store** (`memory_store`) durable facts: preferred lifestyle formats, the influencer's consistent look, vibe/tone, and target platform. Use `kind='preference'` for preferences, `kind='proje