Documentary Film Short · publiczny skill Open Skill

Documentary Film Short

Documentary Film Short to publiczny Open Skill w Digen. Otwórz go, by zobaczyć, co tworzy ten workflow.

@Twinpop

Travel

1 1 2

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1208817/2026-09-25/9pum2yax.jpg

# Documentary Film Short

## When to use
- The user wants a documentary-style short film/clip: observational, naturalistic, "real world" footage of places, people, hosts, environments, or events.
- They give an idea, a one-liner, a script, or reference images, and want a documentary-style short produced from it.
- They want to revise one stage, add/change a scene, or continue/extend an existing documentary short.

## Style rules (always apply unless the user explicitly overrides)
- **Hyperrealistic always**: every frame must read as real, un-staged documentary footage — photorealistic, natural grain, no stylized/cinematic fantasy look, no obvious CGI.
- **No music unless specified**: do not add a musical score/soundtrack by default. Only generate music if the user explicitly asks for it.
- **Subtle ambient sound by default**: add a quiet, natural environmental soundscape (room tone, wind, waves, city ambience, birdsong, etc.) matching the scene by default, unless the user opts out. This is ambient/environmental sound, not a musical score.
- **No dialogue unless specified**: do not write or voice dialogue/narration by default. Only generate TTS narration/voiceover if the user explicitly asks for it.
- **Real skin texture on any host**: any on-camera host/person must show natural, realistic skin texture (pores, subtle blemishes, natural tone) — never airbrushed, plastic, or doll-like.
- **Natural lighting and environmental movement**: light should look like real ambient/source light (window, sun, practicals), and the environment should feel alive with subtle natural motion (leaves, dust, water, people passing, camera drift) — not static or sterile.

## Steps
1. Clarify the subject and scope: what is the documentary about, who/what is on camera (host, subject, location), how long (default a short ~10–30s clip unless they say otherwise), and orientation (default `portrait` 9:16; use `landscape` 16:9 only if they explicitly want widescreen). If they gave a script, use it; if not, draft a short observational treatment.
2. If reference images are provided (a host, location, or subject), use them to keep identity consistent. For a host/subject, generate a clean reference still first if needed.
3. Generate the footage:
   - **From reference image(s) of a host/subject/location** → `generate_video` with `model='ref2v'`, passing the reference image(s) in `input_urls` (keeps identity consistent). Use `name` = scene/shot title, `entity_type` = `storyboard_videos`.
   - **From a single still you want to animate** → `generate_video` with `model='i2v'`, passing that still as the start frame. Use `name` = shot title, `entity_type` = `storyboard_videos`.
   - **Pure text-to-video (no image)** → `generate_video` with `model='t2v'` explicitly (no input image). Use `name` = shot title, `entity_type` = `storyboard_videos`.
   - In every video prompt, restate the style rules: hyperrealistic documentary footage, natural ambient lighting, subtle natural environmental motion, and (if a person is on camera) realistic skin texture.
4. Only if the user asked for narration/voiceover: `generate_tts` with `model='tts.design'` (text + `voice_instruction`) or `model='tts.clone'` (text + `reference_audio_url` for a cloned host voice). Use `name` = narration title, `entity_type` = `voiceovers`.
5. By default, generate a subtle ambient/environmental soundscape: `generate_music` with a prompt describing quiet natural scene ambience (e.g. wind, waves, room tone, city hum, birdsong) matching the footage. Use `name` = ambient track title, `entity_type` = `music`. Skip this only if the user opts out of sound.
6. Only if the user asked for a musical score: `generate_music` with a prompt describing a subtle, documentary-appropriate score. Use `name` = track title, `entity_type` = `music`.
7. If multiple clips (and/or narration/music/ambient sound) need to be combined into one short: `mix_clips` to stitch them in order, adding any audio tracks. Use `name` = final cut titl