Environmental Short Film · عوامی Open Skill

Environmental Short Film

Nature and environmental

@Twinpop

Other

2 2 2

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1208817/2026-09-25/yarmajhc.jpg

# Environmental Short Film

## When to use
- The user wants a cinematic short film that is **environmental only**: pure nature / landscape / weather / atmosphere footage — no people, no characters, no dialogue, no story beats. Typically a series of **3 scenes** (default; user may ask for more or fewer).
- The user wants to **redo/restyle one scene**, **add a scene**, or **stitch the finished scenes into a single film**.
- The user uploads reference images (a location, a mood, a color palette) and wants the scenes to follow them.

## Reference files
- `references/prompt-templates.md`: the exact image and animation prompt templates to fill in for every scene. Read it before writing any prompt.

## Core principles
- **Environmental only.** Never introduce people, characters, faces, animals as subjects, or dialogue. Secondary elements may include water, plants, structures, and (optionally) distant wildlife only as atmosphere — never as the subject.
- **No music, no voiceover, no narration.** The final film is silent ambient footage. Do not call `generate_music` or `generate_tts`.
- **Photorealistic, hyper-realistic, cinematic.** Every prompt ends with the quality/closure lines from the template.
- **Smooth, continuous camera** — no cuts, no flicker, steady even lighting.
- **3 scenes by default.** If the user didn't specify a count, use 3. Confirm the concept (location, time of day, mood) before generating if it's unclear.

## Steps
1. **Clarify the concept** (one short round): what environment/location, what time of day, what mood/color palette, and how many scenes (default 3). If the user uploaded reference images, call `recognize_image` on each to extract location, lighting, palette, and mood, and reuse those cues in the prompts.
2. **Plan with `write_todos`**: list the per-scene image + animation steps and the final stitch, and update the list as you progress.
3. **For each scene (default 3), generate the still image** — read `references/prompt-templates.md`, fill the **image template** for that scene, and call:
   - `generate_image(prompt=<filled image template>, model='t2i', orientation='landscape', resolution='1080P', name='Scene N still', entity_type='storyboard_images')`
   - If the user gave reference images, use `model='i2i'` with `input_urls=[reference asset ids]` instead, keeping the same `name`/`entity_type`.
4. **For each scene, animate the still** — fill the **animation template** for that scene and call:
   - `generate_video(prompt=<filled animation template>, model='i2v', input_urls=[<that scene's still asset id>], orientation='landscape', resolution='1080P', name='Scene N clip', entity_type='storyboard_videos')`
5. **Stitch the scenes into one film** (offer this; do it if the user wants a single deliverable) — call:
   - `mix_clips(clips=[{video: {url: <scene 1 clip>}}, {video: {url: <scene 2 clip>}}, {video: {url: <scene 3 clip>}}], name='Environmental short film', entity_type='final_videos', orientation='landscape')`
6. **Reply** in natural language: list the scenes produced, note that the film is silent ambient footage (no music/voiceover by design), and call `set_guidance` to suggest next actions (e.g. redo a scene, add a scene, change the mood).

## Notes
- Keep prompts in English. Fill every bracketed field of the template; never leave placeholders.
- If a generation fails or looks wrong, report it honestly and retry with an adjusted prompt — never fabricate a URL or claim success.
- Intermediate stills and clips will appear on the canvas alongside the final film; that's expected.
- Politely decline requests that add people, characters, dialogue, or any non-environmental subject — this skill is environmental-only by design.