Prefect Pony XL LoRA vs Prefect Pony XL: which to pick?

Pick Prefect Pony XL LoRA when you have a LoRA file or want the guidance dial. Pick Prefect Pony XL when the checkpoint's own look is enough, since it is the simpler request and has no LoRA slot. Both draw from the same checkpoint family, and both bill per image.

Which LoRAs work here?

LoRAs trained for Pony-family SDXL models, passed as a direct HTTPS link to the .safetensors file. A LoRA for another base loads without an error and changes nothing, so run a with-and-without test before you spend a batch.

Can I use uncensored LoRAs?

Yes. The checkpoint is not filtered and neither is the output. Our strongest content probe rendered as written, and loading a LoRA did not change that. Rights and consent for everything you generate stay with you.

What scale should I use?

Start between 0.7 and 1.0, the range suggested for this checkpoint. In our test a pixel-art LoRA was a lighter touch at 0.5 and the full style at 1. Scale 0 to 4 is accepted; when stacking, lower each one a little.

Do I need the trigger word?

For the style LoRAs we tried, yes: each changed the look when its trigger started the prompt. Read the model card for the exact word and test with and without it.

What guidance scale should I use?

Stay near the default of 3.5 and move in small steps. In our runs 1 looked washed out and soft, 7 gave more contrast and saturation and 12 looked burnt. The range is 1 to 20, because 0 made the model ignore the prompt.

How fast is it, and how is it billed?

In 47 runs, model time was 1.6 to 23.8 seconds with a median of 12.3. Billing is per image at one flat rate: the aspect ratio, the number of LoRAs and the output format do not change it. Current prices are on the model page.

Why did my LoRA request fail?

Usually the link needs a login, since a gated Hugging Face file answers with HTTP 401, or it points at a repository page instead of the file, which asks for the direct download link. Use a public link to the .safetensors file itself. Both are reported as an invalid request.

Prefect Pony XL LoRA API: Uncensored NSFW Anime Images

prefect/pony-xl-lora/text-to-image

Anime and illustration art on Prefect Pony XL, a Pony-lineage SDXL checkpoint, with up to three of your own LoRAs per request, each with its own scale. Score-tag prompts, nine native SDXL aspect ratios, guidance from 1 to 20, and no filter on prompts, weights or results.

From $0.024/image

Run model View API

Input

Text to image

Prompt540/5000

Aspect ratio

Style LoRAs1/3

Art Deco Inspired

Prompt adherence

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

Animate

Edit

Split into layers

{
  "model": "prefect/pony-xl-lora/text-to-image",
  "input": {
    "aspect_ratio": "3:2",
    "loras": [
      {
        "path": "https://huggingface.co/EarthnDusk/Loras2024/resolve/main/Art_Deco_Inspired_Pony_XL.safetensors",
        "scale": 0.7
      }
    ],
    "guidance_scale": 5,
    "output_format": "jpeg",
    "seed": 124,
    "prompt": "end_artdeco_inspo, score_9, score_8_up, score_7_up, source_anime, rating_safe, 1girl, 1boy, adult couple, mature female, mature male, 35 years old, detailed background, scenery, inside a bank vault, giant round vault door standing open, stacks of gold bars on shelves, red security laser beams, she wears a backless gold sequin gown and long black gloves, he wears a black tuxedo jacket worn open with a loosened bow tie, stubble, standing face to face, her hand on his lapel, lips almost touching, geometric gold patterns, dramatic shadows"
  }
}

Introduction

Prefect Pony XL LoRA Text to image draws one anime or illustration image from a tag prompt with up to three of your own LoRA files loaded. It runs Prefect Pony XL, a community SDXL checkpoint merge from the Pony Diffusion lineage, and adds a LoRA slot, a guidance control and nine native SDXL aspect ratios. The platform adds no content filter to the prompt, the weights or the image.

Prefect Pony XL without a LoRA slot has its own page: Prefect Pony XL.

Key features & highlights

Pony-family LoRAs by direct link

The Pony ecosystem has a deep library of style and character LoRAs. Pass the direct link to a .safetensors file and a scale; nothing is uploaded to us in advance, and the file is not reviewed.

Three at once

Up to three LoRAs per request, each with its own scale. In our test three stacked at 0.8, 0.5 and 0.3 loaded together and the main style still read clearly.

Native SDXL shapes

Nine aspect ratios, each mapped to a native SDXL size from 1024 x 1024 to 1536 x 640. In our tests every one came back at exactly the requested size.

Uncensored, with or without weights

Our strongest content probe rendered as written on the first try, without a LoRA and with one loaded. There is no content flag and no second review of the result.

Request parameters

Request parameters for Prefect Pony XL LoRA · Text to Image
Parameter	Required	Description
`prompt`	Yes	Comma-separated tags work best: score tags, a source and rating tag, the head count, then the subject. Put LoRA trigger words first. at least 1 characters · at most 5000 characters
`aspect_ratio`	No	Nine native SDXL shapes from 21:9 to 9:21; 1:1 (1024 x 1024) by default. The ratio changes the shape, not the price. `1:1` · `16:9` · `9:16` · `4:3` · `3:4` · `3:2` · `2:3` · `21:9` · `9:21` · Default: `1:1`
`loras`	No	Up to three {path, scale} entries trained for Pony-family SDXL checkpoints; scale 0 to 4, 1 is neutral, 0.7 to 1 is a good start. Use a public direct link to the .safetensors file. Default: `[]` at most 3 items
`guidance_scale`	No	How closely the image follows the prompt, from 1 to 20; 3.5 by default. In our tests 7 added contrast and 12 burnt the colours. `≥ 1` · `≤ 20` · Default: `3.5`
`output_format`	No	JPEG by default, or PNG or WebP. The format does not change the price. `jpeg` · `png` · `webp` · Default: `jpeg`
`seed`	No	Keeps a set of images comparable while you change one LoRA scale or tag at a time. It does not guarantee the exact same pixels on every run. `≥ 0` · `≤ 2147483647`

Pricing

$0.024 / image

Billed per image. The aspect ratio, the output format and loading LoRAs do not change the price.

Use cases

A house style for a whole set

Load a style LoRA, keep the trigger words at the front and run a list of prompts. Every image in the set shares the look.

Consistent characters

Keep a character LoRA and a seed fixed, then change only the outfit, place and light between prompts.

Uncensored anime for your platform

Load the LoRA your audience expects and write the picture directly in tags. Nothing is filtered on the way in or out.

Explore key use-case tutorials

Prompt guide & pro tips

Score tags first

Start with score_9, score_8_up, score_7_up, then a source tag such as source_anime, a rating tag, the head count (1girl, 1boy) and the subject. Put each LoRA's trigger words at the very front.

Write the age

State that every character is an adult (mature female, adult man, 30 years old). Every example on this page does.

Compare against no LoRA

Run the same prompt and seed without the LoRA. If the two images are identical, the file was trained for another base model: such LoRAs load without an error and do nothing.

Guidance around 3.5 to 7

The default 3.5 is balanced. In our tests 7 gave more contrast and saturation, 12 burnt the colours, and 1 came out washed out.

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": "prefect/pony-xl-lora/text-to-image",
  "input": {
    "aspect_ratio": "1:1",
    "output_format": "jpeg",
    "prompt": "A ceramic teapot on a linen cloth, soft window light, muted palette"
  }
}'

# 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.

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

Bring your Pony LoRAs and a guidance dial to one anime checkpoint

Prefect Pony XL LoRA loads up to three Pony-family LoRA files onto Prefect Pony XL, a community SDXL merge tuned for anime and illustration, and adds a guidance scale you can turn from 1 to 20. In 47 runs, model time was 1.6 to 23.8 seconds with a median of 12.3. Nothing filters the prompt or the image: our strongest content probe rendered as written, with and without a LoRA loaded.

Open the playground

Bring your Pony LoRAs and a guidance dial to one anime checkpoint

LoRAs per request: Up to 3, each {path, scale}; path is a direct HTTPS link to a .safetensors file trained for a Pony-family SDXL model
Scale: 0 to 4; start around 0.7 to 1.0 and lower each one a little when stacking
Guidance scale: 1 to 20, default 3.5; 1 looked washed out, 7 more contrasty, 12 burnt
Speed: 1.6 to 23.8 seconds per image in 47 runs, median 12.3
Price: Billed per image at one flat rate; size, number of LoRAs and output format do not change it
Content filter: None; our strongest content probe rendered as written, with and without a LoRA

Trigger words: Every style LoRA in our tests wanted its wordThe Flat Vector, InkPunk and Art Deco LoRAs each changed the look when their trigger sat at the start of the prompt: vztdzsxx, end_inkpunk_style and end_artdeco_inspo. Put the trigger first, then the quality tags, then the scene.
Wrong base model: A LoRA for another model fails silentlyWe loaded a Z-Image LoRA and got no error and no effect: the picture was pixel-identical to a run without it. If the same prompt and seed look the same with and without the file, it was trained for a different base model.
Limits: What it will not doA fourth LoRA, a scale above 4 or a guidance above 20 is rejected. One request returns one image, and there is no image input, negative prompt or step control. The link must be public and point straight at the file: a gated repository answers with HTTP 401 and a page link asks for the direct download.

Working with LoRAs on Prefect Pony XL LoRA

Load a LoRA with a direct link

Each entry in the loras array is a {path, scale} pair, and the path is a direct HTTPS link to the .safetensors file. A pixel-art LoRA at scale 1 turned the whole frame into pixel art in our test and at 0.5 it was a lighter touch, so the scale is a real dial. The file is fetched when the request runs.

Put the trigger first

Each style LoRA we tried wanted its own word: vztdzsxx for Flat Vector, end_inkpunk_style for InkPunk, end_artdeco_inspo for Art Deco. Read the file's model card, put the trigger at the start of the prompt, then add score_9, score_8_up, score_7_up and a rating tag such as rating_safe.

Turn guidance in the right range
Guidance 1 came out washed out and soft, 3.5 was balanced, 7 gave more contrast and saturation, and 12 looked burnt. Guidance 0 ignored our prompt entirely and drew an unrelated sketch, which is why the public range starts at 1. Raise it in small steps from 3.5 when a LoRA's look needs more bite.
Stack up to three
Three LoRAs at 0.8, 0.5 and 0.3 worked together and the main style stayed readable. Keep the main style highest, bring the others in lower and raise them in small steps; when you stack, lower each scale a little compared with running it alone.
Stay on the nine ratios
The nine aspect ratios map to SDXL-native sizes, from 640×1536 for 9:21 to 1536×640 for 21:9, and every one completed at exactly the requested size. When we pushed past the native buckets, 1536×864 drew duplicated people in 3 of 4 tries, so the public API only offers the native shapes.

Who uses Prefect Pony XL LoRA

Comic and illustration artists
An ink look across a whole series
With the InkPunk LoRA and end_inkpunk_style in front, we drew a woman in a leather jacket leaning on a motorcycle at 0.9 and a bearded man at a desert gas station at 0.8. Change the subject line and keep the LoRA, the scale and the quality tags to hold one look across a set.
Poster and brand illustrators
Art Deco and flat vector, on demand
The Art Deco LoRA gave us a gold-and-black ballroom tango and a flapper on a grand staircase, and the Flat Vector LoRA gave a clean poolside poster. Each came from one request with the trigger first, so a poster set is a handful of calls.
Visual-novel and game teams
A cast in one house style
A woman in a kimono on lantern-lit stairs, a bartender behind a bar and a detective in a long coat share the same quality tags, the same checkpoint and the same rating tag. The detective uses the InkPunk LoRA at 0.8 and the other two use none, which shows how far one checkpoint stretches.
Key-art and wallpaper makers
Wide night scenes that keep the mood
Wide frames work at 16:9 and 21:9, which are 1344×768 and 1536×640. Describe the weather and the light in tags, keep the character in the prompt and leave the framing to the ratio.

From first LoRA to finished set with Prefect Pony XL LoRA

Find a LoRA trained for Pony
Read its model card for the trigger word and a suggested strength, then copy the direct download link to the .safetensors file.
Run it with and without the file
Same prompt, same seed, one run each. If nothing changes, the file was probably trained for another base model.
Ladder the scale
Hold everything else still and try 0.5, 0.8 and 1. Keep the value where the look lands without taking over the subject.
Tune guidance last
Start at the default 3.5 and move it in small steps; the useful range in our runs was roughly 3.5 to 7.
Stack a second or third file
Add each extra LoRA at a lower scale than the main one and raise it in small steps. Then run the set with one seed per frame.

Prefect Pony XL LoRA prompts worth copying

A flat vector poster

With the Flat Vector LoRA at scale 1 and the trigger first.

vztdzsxx, vector art, flat color, vector illustration, score_9, score_8_up, score_7_up, source_anime, rating_safe, 1girl, solo, mature female, adult woman, tanned skin, long black hair, yellow one-piece swimsuit, sunglasses on her head, sitting at the edge of a swimming pool, palm tree shadows, summer afternoon

A dance in Art Deco gold

With the Art Deco LoRA at scale 1 and guidance 5; the wide 16:9 ratio suits the ballroom.

end_artdeco_inspo, score_9, score_8_up, score_7_up, source_anime, rating_safe, 1girl, 1boy, adult couple, mature female, mature male, dancing a tango in a 1920s ballroom, she wears a gold beaded dress with a high slit, he wears a black tuxedo, dramatic dip, chandeliers, geometric gold patterns

An ink-wash portrait

With the InkPunk LoRA at scale 0.8 and the 3:4 ratio.

end_inkpunk_style, inked, ink wash, ink splatter, score_9, score_8_up, score_7_up, source_anime, rating_safe, 1boy, solo, mature male, adult man, 35 years old, short beard, black leather jacket, white t-shirt, leaning against a vintage car at a desert gas station at dusk, orange sky, looking at viewer

All Prefect Pony XL models2

Prefect Pony XL LoRACurrentText to image$0.024per image
Prefect Pony XLText to image$0.015per image

LoRA guide

Best NSFW LoRAs for Pony · SDXL

Workbench