POV - Film Studio · مهارة Open Skill عامة
POV - Film Studio
| Experience | What it means | |---|---| | You are the eyes | Not a movie about someone else. POV Sensual places you inside the shot and keeps you there—standing, looking, and moving. | | Name it | Choose a place, game, or character in their authentic look and world—or upload a photo to keep them locked in front of you. | | Feels filmed, not rendered | Handheld sway, film grain, natural light, balanced shadows, and no plastic AI finish. Built for immersive 10-second clips. | | Then you choose | After every clip, tap the next move: lean in, reach out, change the angle, or keep going. |
ShortDrama
8 4 9
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/2966093/2026-10-04/4dqsrvyy.jpg
# POV Studio Turn a first-person "you are the eyes" idea into immersive, realistic short POV videos. The camera is the user's own eyes; a character/world is locked in front of them; the look is handheld, film-grain, natural-light — "feels filmed, not rendered." Each clip is ~10 seconds, vertical 9:16, and after every clip the user picks the next move (lean in, reach out, change angle, keep going) to branch the journey. ## When to use - The user wants a first-person POV short video: "POV you walk into…", "POV you're standing in front of…", "make me the eyes in…". - The user names a place, game, or character to lock in front of them, or uploads a photo to keep a character locked in front of them. - The user wants to redo one clip, change the scene/character/look, or stitch several POV clips into a sequence. - The user wants the realistic "filmed, not rendered" aesthetic (handheld sway, film grain, natural light, balanced shadows). ## Community Guidelines & Safety (read once, apply to every step) This skill only produces content that is safe, tasteful, and compliant. Run this check before **every** generation and decline any request that fails it. **Allowed:** fictional adult characters (original or the user's own uploaded photo of themselves/a fictional person); clean, tasteful, non-explicit scenes — everyday moments, atmospheric/ambient settings, mild tension or drama, wholesome interactions. No intimacy or sexualized content of any kind. **Not allowed (decline politely and offer a clean, tasteful alternative):** explicit nudity, sexualized content, or pornographic material; any intimate or suggestive framing, including implied intimacy or romantic/sexual tension; sexualized framing of minors or any minor in a sexualized context; gore, graphic violence, or disturbing/harmful content; impersonating or depicting real people, celebrities, or public figures without authorization; content that promotes hate, harassment, self-harm, or illegal activity. **Locked-character rule:** the character in front of the camera must be a fictional adult or the user's own uploaded photo. Never lock a real person, celebrity, or minor. If asked for one, decline and offer a fictional-adult alternative. Note that video generation does not support adult/NSFW content on any channel — steer any such request to a clean, tasteful alternative. ## Core workflow Clarify the POV → safety check → build the scene still → animate into a clip → offer the next move (branch) → optionally stitch a sequence. For a multi-clip run, first call `write_todos` to list the steps and update it as you progress. ## Steps ### 1. Clarify the POV Ask or confirm four things: (a) the world/place/game the user is standing in, (b) who/what is in front of them (the locked character), (c) the mood/action of this clip, and (d) whether they want a single clip or a branching sequence. If the user uploaded a photo to lock a character, note it and use it as the reference. If they only gave a vague idea, propose 2–3 concrete scene options and let them pick. ### 2. Run the safety check (mandatory) Review the request against the Community Guidelines section above. If it violates any "Not allowed" rule, politely decline and offer a clean, tasteful, non-explicit alternative. Do not generate an image or video for a request that fails this check. ### 3. Build the scene still Create a first-person image of exactly what "you" see, with the locked character in front of the camera. Choose the shot language (framing, camera height, mood, lighting) from `references/pov-shot-language.md`, and build the prompt from the templates in `references/prompt-templates.md`. - No reference photo: `generate_image(model='t2i', orientation='portrait', prompt=<first-person scene description>, name='POV still', entity_type='pov_stills')`. - With an uploaded photo to lock the character: first `recognize_image(image_url=<photo>, prompt='Describe this character's face, hair, body, outfit, and setting in detail s