Is P-H3-Image the original H3 image generator?

Pruna AI adapted MiniMax H3 open weights into a separate still-image model, released on November 3, 2026. In our tests the original H3 image generator and P-H3-Image produced different pictures from a matching prompt and seed. See [MiniMax H3 Image LoRA](/models/minimax-h3-image-lora) for the original H3 still-image option at 1k or 2k.

What does the 4k setting mean?

It selects a tier of about 16.8 megapixels, with dimensions determined by the ratio. In our tests a 16:9 generation measured 5472 x 3072 and an edit following its source measured 3360 x 5024.

How should I order pictures for editing?

Start with the scene you want to retain, then add the references that supply replacements. In the stage test, our instruction retained the stage and green dress from while taking the person from .

Is P-H3-Image uncensored for NSFW images?

There is no content filter, and our tests recorded no refusals. Extreme requests can return toned down; every person you describe must be an adult.

What timing should I allow?

In our tests 1k model time was usually 5–17 seconds, with occasional runs over 30 seconds. One 4k generation took 241 seconds and one 4k edit took 52 seconds; these measurements exclude queueing and transfer.

Which settings affect the bill?

The resolution tier sets the per-image charge. Editing adds a small fee for every input beyond the first; aspect ratio and file format leave the charge unchanged. Failed requests are not billed.

Where do I add an H3 LoRA?

Use [P-H3-Image LoRA](/models/p-h3-image-lora) to attach MiniMax H3 LoRA files to generation or editing. The base page covers the model without those attachments.

P-H3-Image API: NSFW Images, 4K Tiers, 9-Image Edits

pruna-ai/p-h3-image/text-to-image

Write a scene and receive a single P-H3-Image still in one of five pixel tiers, from 1k through 4k.

From $0.024/image

Run model View API

Input

Text to image

Prompt245/5000

Aspect ratio

Resolutionaffects price

Output format

JPEGPNGWebP

Seed

Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Example result

Edit

Split into layers

Animate

{
  "model": "pruna-ai/p-h3-image/text-to-image",
  "input": {
    "aspect_ratio": "2:3",
    "resolution": "4k",
    "output_format": "png",
    "seed": 7,
    "prompt": "Editorial portrait of a woman in her thirties with short platinum hair and a charcoal wool coat, standing on a windy concrete rooftop at dusk, city lights soft behind her, shot on medium format film, natural skin texture, shallow depth of field."
  }
}

Introduction

P-H3-Image Text to Image turns your scene description into one still. Pruna AI released its open weights on November 3, 2026, adapting MiniMax H3 for image generation and editing.

Key features & highlights

Select 1k, 1.5k, 2k, 3k or 4k independently of the 15 aspect ratios. In our tests the 4k setting at 16:9 produced 5472 x 3072 pixels. The result can be JPEG, PNG or WebP.

Request parameters

Request parameters for P-H3-Image · Text to Image
Parameter	Required	Description
`prompt`	Yes	Required scene description; there is no default text. Describe clothing, setting and light concisely rather than adding a long adjective list. at least 1 characters · at most 5000 characters
`aspect_ratio`	No	Choose among 15 shapes: 1:1, 1:2, 2:1, 1:3, 3:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 9:21, 21:9. `1:1` · `16:9` · `9:16` · `4:3` · `3:4` · `3:2` · `2:3` · `21:9` · `9:21` · `2:1` · `1:2` · `3:1` · `1:3` · `5:4` · `4:5` · Default: `1:1`
`resolution`	No	Select 1k (default, about 1 MP), 1.5k (~2.4 MP), 2k (~4 MP), 3k (~9.4 MP) or 4k (~16.8 MP). This field determines the billed size tier. `1k` · `1.5k` · `2k` · `3k` · `4k` · Default: `1k`
`output_format`	No	The file encoding defaults to jpeg; png and webp are the other supported values. Choosing an encoding adds no fee. `jpeg` · `png` · `webp` · Default: `jpeg`
`seed`	No	Optional random-source setting with no required fixed value. Reuse it while changing one setting for a fair comparison; identical output is not guaranteed. `≥ 0` · `≤ 2147483647`

Pricing

Each generated image is billed at its selected size tier: 1k, 1.5k, 2k, 3k or 4k. Ratio and format carry no surcharge. In editing mode, every source beyond the first adds a small charge. Failed requests are not charged.

Pricing
Specification	Price
`1k`	$0.024 / image
`1.5k`	$0.036 / image
`2k`	$0.048 / image
`3k`	$0.096 / image
`4k`	$0.144 / image

Use cases

Use scene descriptions for editorial portraits, diner film stills or short neon signs. For changes to existing pictures, choose Image Editing.

Explore key use-case tutorials

Prompt guide & pro tips

Name a subject, clothing, place and light source. Pruna AI says short quoted words can render, but long passages and small lettering can become false text. Keep the seed while changing one element to compare results; it does not lock the output. In our tests larger tiers took longer, with one 4k generation reaching 241 seconds of model time.

API quick start

# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
  -d '{
  "model": "pruna-ai/p-h3-image/text-to-image",
  "input": {
    "prompt": "A ceramic teapot on a linen cloth, soft window light, muted palette",
    "resolution": "1k",
    "aspect_ratio": "1:1",
    "output_format": "jpeg"
  }
}'

# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $SPICY_API_KEY"

See the API reference for authentication, the complete schema, polling, and error handling.

Missing the model you need?

PRUNA AI IMAGE API

P-H3-Image API at a glance

Five size tiers for a written scene or an edit built from numbered pictures.

Largest tier: 4k; About 16.8 megapixels at the 4k setting.
Size tiers: 5; The intermediate settings include 1.5k and 3k.
Editing inputs: 9; One picture to edit, with room for eight reference pictures.
Ratio choices: 15; Select a square, tall or wide composition without a ratio surcharge.
Image per request: 1; Both modes return a still rather than video or audio.

Results

What you can make with P-H3-Image

Size selection
A rooftop portrait requested at 4k
The rooftop example uses the 4k tier and a 2:3 ratio. Size is selected independently of the scene description.
4k request, 2:3; no LoRA.
Quoted words
Three words on a neon sign
The prompt quotes OPEN ALL NIGHT and places the lettering above a green door. Pruna AI says short quoted phrases can render, but long or small text can break down.
1k, 1:1; no LoRA.
Source order
Keep the stage and replace the performer
The stage occupies the first input slot, with the replacement person in the second. Our test kept the microphone, lighting and green dress while changing the performer.
1k, 3:4; two input images, no LoRA.

Content test

Is P-H3-Image uncensored? What we tested

Content levels tested on P-H3-Image, by mode
Level	Text to image	Image editing
Mild	RendersTested Nov 5, 2026	Not tested yet
Moderate	RendersTested Nov 5, 2026	RendersTested Nov 5, 2026
Strong	RendersTested Nov 5, 2026	RendersTested Nov 5, 2026
Very strong	RendersTested Nov 5, 2026	RendersTested Nov 5, 2026
Maximum	SoftenedTested Nov 5, 2026	Not tested yet

How we test

How each mode behaves

All modes: Pruna AI says lower tiers show more detail artifacts and that long, adjective-heavy prompts can hurt results.
All modes: There are no negative-prompt, guidance or step controls, and a request cannot generate a batch.
All modes: Use this model responsibly: every person you describe must be an adult. You are responsible for what you generate and publish.
All modes: Generated images are kept 14 days, then deleted.

Use cases

What to build with the P-H3-Image API

Editorial scenes
Describe a rooftop at dusk or a rain-lit diner with specific clothing and light sources.
Short signs
Put a few words in quotation marks and name the sign that should display them. Pruna AI says short quoted text can work, while small lettering and long passages can become false text.
New portrait settings
Give the editor one source and describe the replacement outfit and location. In our tests the person remained recognizable after these changes.
Replace a stage performer
Keep the stage image in the first slot and identify the replacement person with a second picture.

Prompts

P-H3-Image prompt examples you can copy

Text to image
Rooftop at dusk
Editorial portrait of a woman in her thirties with short platinum hair and a charcoal wool coat, standing on a windy concrete rooftop at dusk, city lights soft behind her, shot on medium format film, natural skin texture, shallow depth of field.
Why it works. The prompt sets a rooftop, dusk lighting and a film treatment in a single scene description.
Try this prompt
Text to image
Rainy diner booth
Film still of a man and a woman in their forties sharing a booth in a 1970s roadside diner at night, rain on the window, red vinyl seats, coffee cups, warm tungsten light inside and blue neon outside, 35mm film grain.
Why it works. Warm interior lighting and blue neon outside give the diner scene two clearly named light sources.
Try this prompt
Text to image
Neon alley door
A narrow alley doorway at night with a hand-bent pink neon sign reading "OPEN ALL NIGHT" above a green door, wet cobblestones reflecting the glow, light mist, photographic.
Why it works. Quotation marks isolate the three requested sign words from the surrounding scene description.
Try this prompt
Image editing
Stage performer replacement
Replace the singer in <Picture 1> with the woman from <Picture 2>. Keep the stage, microphone, spotlight and green velvet dress from <Picture 1>; keep the face and hair of <Picture 2>.
Why it works. Numbered tokens assign the retained stage to the first input and the replacement person to the second.
Try this prompt
Image editing
Red gown by the sea
<Picture 1> is a woman. Keep her face and hair. Change her outfit to a deep red satin evening gown and place her on a sunlit stone terrace above the sea at golden hour.
Why it works. The instruction names what to preserve before specifying the new gown and seaside location.
Try this prompt

Pricing

P-H3-Image API pricing

One image starts at $0.024 at 1K and $0.144 at 4K.

See every price tier How pricing works

Get started

How to use the P-H3-Image API in three steps

Step 1
Describe a scene in a few sentences
Write the subject, clothing, location and lighting in prompt. Pruna AI says adjective piles and overlong descriptions can reduce quality.
Step 2
Choose size separately from shape
Set resolution for the pixel tier and aspect_ratio for the composition. The size tier affects billing; the chosen ratio does not.
Step 3
Put the scene to preserve first
For an edit, place the picture whose scene should remain first in image_urls. Use <Picture 1> for that source and <Picture 2> for the next reference.

Open the playground Get an API key

Parameter reference and code samples are in the API and Readme tabs at the top of this page.

FAQ

P-H3-Image API FAQ

About P-H3-Image

Choose P-H3-Image for still generation or numbered-reference editing with five pixel tiers. Begin with a concise instruction, choose the output shape, and raise the resolution when you need a larger result.

All P-H3-Image models2

P-H3-ImageCurrentText to image · Image editingFrom $0.024per image
P-H3-Image LoRANewText to image · Image editingFrom $0.042per image

Workbench

Input

Output

Introduction

Key features & highlights

Request parameters

Pricing

Use cases

Prompt guide & pro tips

API quick start

A rooftop portrait requested at 4k

Three words on a neon sign

Keep the stage and replace the performer

How each mode behaves

Editorial scenes

Short signs

New portrait settings

Replace a stage performer

Rooftop at dusk

Rainy diner booth

Neon alley door

Stage performer replacement

Red gown by the sea

Describe a scene in a few sentences

Choose size separately from shape

Put the scene to preserve first

Is P-H3-Image the original H3 image generator?+

What does the 4k setting mean?+

How should I order pictures for editing?+

Is P-H3-Image uncensored for NSFW images?+

What timing should I allow?+

Which settings affect the bill?+

Where do I add an H3 LoRA?+

About P-H3-Image

All P-H3-Image models2

Related models and next tasks