P-Image-Upscale

P-Image-Upscale is Pruna’s model for general quality improvements on generated or edited images.

It works especially well as a final polish pass when the composition is already right and you want a cleaner, sharper, more finished result without changing the core scene.

Tip

For an end-to-end walkthrough covering both p-image and p-image-edit workflows, see Use P-Image-Upscale for General Quality Improvements and Final Polish.

What it’s good for

• General quality improvements on existing images that already work compositionally.

• Sharper details, cleaner edges, and stronger final presentation.

• Clearer text-heavy visuals such as ads, posters, and promo creatives.

• Polished product, retail, and stylized creative outputs that need a more publishable final image.

• Final enhancement after p-image or p-image-edit when content decisions are already done.

What it’s not for

• Major content changes or scene redesigns.

• Fixing weak composition, poor prompt choice, or missing subject detail.

• Replacing editing workflows when you really need p-image-edit.

Comparisons

The examples below show general quality improvements on images that already work. The goal is not to redesign the content, but to make the final image feel cleaner, sharper, and more publishable.

General detail cleanup — Drag to compare

↔ Drag starting image p-image-upscale

Stronger final finish — Drag to compare

↔ Drag starting image p-image-upscale

↔ Drag starting image p-image-upscale

Retail product polish

↔ Drag starting image p-image-upscale

Stylized art refinement

↔ Drag starting image p-image-upscale

Text clarity

↔ Drag starting image p-image-upscale

Product edge cleanup

↔ Drag starting image p-image-upscale

Catalog image polish

↔ Drag starting image p-image-upscale

Game art cleanup

↔ Drag starting image p-image-upscale

Game texture cleanup

↔ Drag starting image p-image-upscale

Character art cleanup

↔ Drag starting image p-image-upscale

Ad creative polish

↔ Drag starting image p-image-upscale

Promo visual cleanup

↔ Drag starting image p-image-upscale

Poster-style art refinement

↔ Drag starting image p-image-upscale

Scenic art cleanup

When to use it

• Use it for general quality improvements when you already have an image you like.

• Apply it after p-image when the generation is already right and needs a stronger final finish.

• Apply it after p-image-edit once the edit is final and you want a more polished output.

• Use it when you want sharper details, cleaner text, and a more publishable final image without changing the composition.

Integration

P-Image-Upscale is available through Replicate. It supports target resolutions from 1 to 8 MP, with optional detail and realism enhancement for stronger final output quality.

For the API-hosted guide version of this workflow, see https://docs.api.pruna.ai/guides/models/p-image-upscale.

Model version

prunaai/p-image-upscale:7135ff723ecea89c0f67afcd51e4904904586e351093465bdc7beed45941b3e0

Example (Replicate Python client)

import replicate

input = {
    "image": "https://replicate.delivery/pbxt/Omb2L1aOCEjqoP5R1bm6biU0MsbI1PGMtsoUdoIdOaO4Pc16/rltivouao62wpy9zcl5o3woj1wam.jpg",
    "upscale_mode": "target",
    "target": 8,
    "enhance_details": True,
    "enhance_realism": True,
    "output_quality": 50,
    "output_format": "jpg",
}

output = replicate.run(
    "prunaai/p-image-upscale:7135ff723ecea89c0f67afcd51e4904904586e351093465bdc7beed45941b3e0",
    input=input,
)

print(output.url)

with open("output.jpg", "wb") as file:
    file.write(output.read())

Configuration

Required parameters

Parameter	Type	Description
image	file	Input image to upscale.

Optional parameters

Parameter	Type	Default	Description
upscale_mode	string	"target"	Upscale mode. Use target for a fixed megapixel target, or factor to scale each side by a multiplier.
target	integer	4	Target resolution in megapixels when upscale_mode is target.
factor	number	2	Scaling factor applied to each side when upscale_mode is factor. Output is capped at 8 MP.
enhance_details	boolean	false	Enhance fine textures and small details.
enhance_realism	boolean	true	Improve realism. This can deviate more from the original image and is often useful for AI-generated inputs.
output_format	string	"jpg"	Output image format.
output_quality	integer	80	Output quality from 0 to 100. This does not affect PNG output.
disable_safety_checker	boolean	false	Disable the safety checker for generated images.

Tip

See Use P-Image-Upscale for General Quality Improvements and Final Polish for a tutorial that covers p-image -> p-image-upscale and p-image-edit -> p-image-upscale end-to-end.