Police Simulator Video · skill Open Skill công khai
Police Simulator Video
Police Simulator Video là một Open Skill công khai trên Digen. Mở ra để xem workflow này tạo gì.
Anime
8 5 37
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3949831/2026-08-19/tgcf4nz3.png
# Character Simulator Video Maker Turn a person (from an uploaded photo) into a "character simulator game" style short video: game-screenshot-style keyframes (with HUD: mission banner, minimap/status meters, speedometer or vitals, on-screen labels) showing the person as the chosen role (police officer, thief, doctor, astronaut, firefighter, chef, pilot, farmer, …), animated into motion video. Segments are generated one at a time: before each keyframe you plan the content flow and pick the most suitable scene, then generate the keyframe; after each clip the user chooses to continue (next segment) or finish (final video gets background music). The first clip is made from the keyframe via `ref2v` (multi-reference, photo + keyframe) for the strongest character identity consistency; each continuation clip uses the previous clip's final frame as its start frame and transitions toward a newly generated intermediate frame. All on-screen text follows the user's input language. ## When to use - The user wants a simulator-style game video of any character (police simulator, thief/heist, doctor, astronaut, firefighter, chef, pilot, farmer, custom role, …). - The user uploads a photo (their own or someone else's) and wants the person in it turned into a simulated character, e.g. "make me a police officer", "simulate police", "simulate a thief with my photo". - After an image is uploaded, the skill MUST first present role-choice buttons (simulate police / simulate thief / simulate other role) and only continue generating after the user picks one. ## Style guarantee (read first) Every keyframe image AND every generated video must clearly look like a simulator video game — AAA/stylized game render with the game UI HUD overlay (mission banner, minimap, status meters), gameplay-like camera and motion. It must read as gameplay footage, NOT live-action film or photorealistic cinema. State this explicitly in every `generate_image` and `generate_video` prompt (e.g. "clearly a video game screenshot / gameplay footage, not live-action"); if a result drifts toward realism, regenerate with a stronger style prompt rather than accepting it. ## Text-to-image rule (mandatory) Whenever text-to-image is involved (a `generate_image` call with NO input image, i.e. pure text-to-image), the ONLY allowed model/channel is `t2i.hd.lite`. Never use any other text-to-image model or channel in this skill. This does not affect image-to-image keyframes (which use `i2i.hd.lite` with the uploaded photo / previous frame as input). ## Default video duration - The DEFAULT duration for EVERY `generate_video` call in this skill is **10 seconds**: always pass `duration=10` unless the user explicitly asks for a different length. - Each generated clip is 10s by default; when the user finishes, the final merged video is the concatenation of those clips (each 10s) with background music. - If a model/channel auto-downgrades or clamps the duration, that is fine — never silently request a longer duration on your own. - Mention this default to the user if they ask about video length; offer to adjust it as an optional tweak. ## Reference files (read before generating) These files hold the detailed requirements; read the relevant one before each step and follow it: - `references/role-scenarios.md` — role-specific missions, scenes, outfits, props, HUD accents for every supported role (incl. custom roles) and segment arc tips. - `references/hud-style-guide.md` — HUD/game-UI requirements, per-role accent colors, on-screen text language rule, placement/composition rules. - `references/prompt-templates.md` — ready-to-fill prompt skeletons for keyframe, first clip, continuation, and music; also the default 480P resolution rule and the default 10s duration rule. - `references/continuation-guide.md` — detailed Continue/Finish workflow: recognize_video, planning pass, intermediate frame, start/end-frame animation (duration=10), music merging. ## Workflow (one segment at a time) 1. **Role c