p-video-replace

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-replace` in backticks, confirm PRUNA_API_KEY (or stop with signup links from pruna-api), then ask for required inputs. Open intake → generation-diversity clarification intake before the first POST. Draft swaps with Prompt craft (dynamic + faithful) — do not paste skill examples. Redirect when When NOT to use fits better.

Prompt craft (dynamic + faithful)

instruction_prompt maps source slots → reference images → preserve-list. Change only what the user asked; keep camera, audio, and unmentioned elements identical.

Fidelity check (before pay): every stated swap and every stated keep must appear in instruction_prompt. Do not “spice” extras when they asked for one identity swap.

Skill boundary

Use p-video-edit for prompt-driven attribute, environment, object, text, or lighting edits. Use p-video-animate when the user has a still and wants it to perform from a separate template video.

When NOT to use

Use a different skill instead:

HTTP (curl)

Upload source assets

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

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

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

Use each response urls.get in input.video and input.images.

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-replace' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "images": [
        "https://api.pruna.ai/v1/files/reference-person-a-def456",
        "https://api.pruna.ai/v1/files/reference-person-b-ghi789"
      ],
      "resolution": "720p",
      "target_fps": "original",
      "instruction_prompt": "Replace the woman on the left (olive coat) with the first reference. Replace the man on the right (navy jacket) with the second reference. Preserve walking pace, camera tracking, and audio."
    }
  }'

Poll and download: follow pruna-api. Output duration follows the source video.

Complete the random seed ritual from generation-diversity before writing prompts.

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-replace' \
  -H 'Try-Sync: true' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "images": [
        "https://api.pruna.ai/v1/files/reference-image-def456"
      ]
    }
  }'

Before generating

1. Complete Prerequisites guide reading order (generation-diversity → video-prompting).
2. Ritual seed → draft a dynamic + faithful instruction_prompt (section above) → confirm video, images (1–4), resolution, target_fps, instruction_prompt, and swap intent (character / clothing-only / object / mixed).
3. Pruna notes: identity comes from images; slot mapping from instruction_prompt (name source element → map each ref → preserve-list → “only X changes”). Vague targets and tiny on-screen objects drop quality. Runware map: inputs.video → video, referenceImages → images, positivePrompt → instruction_prompt.

Required input

• video (string URL): source RGB .mp4
• images (array of 1–4 string URLs): identity reference(s)

Common optional fields

• resolution: 720p (default) or 1080p
• target_fps: original (default), 24, or 48
• instruction_prompt (string)
• save_audio (boolean, default true)
• seed, disable_safety_checker

Typical next steps

Common follow-ons after this skill: