p-image-upscale

Use when the user wants to upscale image resolution, enhance detail in an existing still, or prepare photos for print, large crops, or quality demos.

Pruna AI@pruna-ai

Install

openclaw skills install @pruna-ai/p-image-upscale

p-image-upscale (Pruna)

AI upscaling with configurable target resolution. Full parameters: p-image-upscale model docs.

Shared HTTP patterns: pruna-api.md (upload, poll, download)

HTTP (curl)

Upload source image

bash

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

Use urls.get as input.image.

Create (async — recommended)

bash

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H "apikey: ${PRUNA_API_KEY}" \
  -H 'Model: p-image-upscale' \
  -d '{
    "input": {
      "image": "https://api.pruna.ai/v1/files/FILE_ID",
      "target": 8,
      "enhance_details": true,
      "output_format": "png"
    }
  }'

Poll and download: pruna-api.md.

Before generating

Generation diversity — ritual seed + axis rotation (optional API seed if supported).
Confirm target MP (1–128), enhance_details / enhance_realism, and output_format with the user so upscale matches destination. Validate outputs with p-image-upscale-quality-checklist.md.

When to upscale

Use case	Typical `target` MP	Notes
Print / billboard / extreme crop	8–128	Confirm cost/latency with user
Mood board / packshot enlargement	4–16	Optional in pruna-generative-pipeline recipes A/B/C
Before/after slider video	`generate_upscale_comparison.py`	Not used in avatar or motion-transfer pipelines

Video workflows (multi-scene-avatar-video, p-video-animate, p-video-replace) feed p-image / p-image-edit outputs directly into video models after the slop gate—do not add an upscale step unless the user explicitly asks for print-scale stills.

Recommended defaults: enhance_details: true, enhance_realism: false. Use enhance_realism: true only when the source is already photoreal and you need extra skin texture—it can add waxy artifacts on synthetic edits.

Prerequisites

image must be a reachable URL (upload via POST /v1/files first if needed).

Required input

image (string URL)

Common optional fields

target: integer megapixels 1–128 (default 4). Model upscales toward this output size; confirm current limits on p-image-upscale model docs.
output_format: jpg, png, webp
output_quality: 0–100 (not used for PNG)
enhance_details, enhance_realism (booleans; realism can drift more from source)
disable_safety_checker

Example: synchronous

bash

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H "apikey: ${PRUNA_API_KEY}" \
  -H 'Model: p-image-upscale' \
  -H 'Try-Sync: true' \
  -d '{
    "input": {
      "image": "https://api.pruna.ai/v1/files/FILE_ID",
      "target": 4,
      "enhance_details": true,
      "enhance_realism": false,
      "output_format": "png"
    }
  }'

Typical next steps

Further edit: upscaled URL → p-image-edit for layout or copy-safe tweaks.
Before/after demo: zoom + slider from any still pair → generate_upscale_comparison.py.
Avatar / motion video (no upscale): multi-scene-avatar-video.

Related workflow

Upscale comparison reels: generate_upscale_comparison.py.