p-video-edit

Prerequisites

Install and load these skills before generating (skip if already in context via @pruna):

Or install the full suite once: npx skills add PrunaAI/pruna-skills@pruna -y

Follow each skill's Before generating / craft sections — do not restate guide content here.

Agent habit

In the first reply, name `p-video-edit` in backticks, confirm PRUNA_API_KEY (or stop with signup links from pruna-api), then ask for the source video (max 15s) and edit brief. Open intake → generation-diversity clarification intake before the first POST. Draft edits with Prompt craft (dynamic + faithful) — do not paste skill examples. Redirect when When NOT to use fits better.

Prompt craft (dynamic + faithful)

prompt is surgical and request-locked. One principal change per run. Describe the desired final state, then name what must stay unchanged. Change only what the user asked; keep camera, motion, audio, and unmentioned elements identical.

Fidelity check (before pay): the prompt must still satisfy the user’s change and every stated keep. Do not “spice” extras when they asked for one attribute or environment edit.

Weak jobs (avoid or warn): a brand-new scene/plot; adding an object with its own independent motion (especially a new in-hand shape); changing camera angle, camera motion, or zoom.

Skill boundary

Use p-video-replace when the job is a character swap from required reference stills. Use this skill for prompt-driven attribute, environment, object, text, or lighting edits.

When NOT to use

Use a different skill instead:

HTTP (curl)

Upload source video

curl -X POST "https://api.pruna.ai/v1/files" \
  -H "apikey: ${PRUNA_API_KEY}" \
  -F "content=@/path/to/source-video.mp4"

Use the response urls.get in input.video. Source video maximum length: 15 seconds.

Optional reference images

curl -X POST "https://api.pruna.ai/v1/files" \
  -H "apikey: ${PRUNA_API_KEY}" \
  -F "content=@/path/to/reference-image.png"

Use each response urls.get in input.images (up to 4; jpg, jpeg, png, webp).

Create (async — recommended)

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H "apikey: ${PRUNA_API_KEY}" \
  -H 'Model: p-video-edit' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "prompt": "Change only the sky to a sunset. Preserve the subject, camera path, and lighting on the foreground."
    }
  }'

Poll and download: follow pruna-api. Output duration follows the source video. Billing is per returned second ($0.045 standard, $0.025 draft). Rate limit: 50 requests per minute.

Complete the random seed ritual from generation-diversity before writing prompts — do not pass the ritual string as API seed. Optional seed only when the user requests reproducibility.

Create (sync — quick test only)

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H "apikey: ${PRUNA_API_KEY}" \
  -H 'Model: p-video-edit' \
  -H 'Try-Sync: true' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "prompt": "Change only the sky to a sunset. Preserve the subject, camera path, and lighting on the foreground.",
      "draft": true
    }
  }'

Before generating

1. Complete Prerequisites guide reading order (generation-diversity → video-prompting).
2. Ritual seed → draft a dynamic + faithful prompt (section above) → confirm video (≤15s), prompt, optional images (0–4), draft, save_audio, and edit intent (attribute / remove / add / environment / relight / text / reference-guided).
3. Pruna notes: one principal change per run. Identity or product match comes from optional images; the edit instruction lives in prompt. Vague targets and brand-new scenes drop quality. Do not send a task field — it is not in the public API.

Required input

• video (string URL): source video to edit. Maximum length: 15 seconds
• prompt (string): text instruction describing the edit

Common optional fields

• images (array of up to 4 string URLs): reference image(s) for reference-guided editing (jpg, jpeg, png, webp)
• prompt_upsampling (boolean, default true)
• draft (boolean, default false): faster, lower-quality preview; billed at $0.025 per output second
• save_audio (boolean, default true)
• seed (integer): set only for reproducible reruns

Typical next steps

Common follow-ons after this skill: