Will it render the strongest NSFW content?

Without a matching LoRA, often not: the base model softened most prompts from the Strong level up in our tests. LoRA files and results are not filtered, but no LoRA for those levels on FLUX 3 Klein was public when we tested. For the highest levels without a LoRA, use the distilled FLUX 3 Klein LoRA.

Which LoRAs work here?

LoRAs trained on FLUX 3 Image Klein. Black Forest Labs recommends training on Base, and in our tests a character LoRA and an outpainting LoRA trained on Base both changed the result. A FLUX.2 [klein] 9B LoRA loaded and changed nothing, pixel for pixel.

My LoRA loaded but nothing changed. Why?

Usually the file was trained for another model, such as FLUX.2 [klein]. A camera-angle LoRA saved in diffusers format also made no visible difference in our tests. None of these raise an error, so compare the same seed with and without the file.

Why did my LoRA request fail?

The link has to download the .safetensors file itself, publicly, and the file must be under 2 GB. A model page link, a gated file or a larger file fails with a LoRA loading error. Failed requests are not charged.

Will a LoRA trained here run on the distilled Klein?

Black Forest Labs says a LoRA trained on Klein Base loads into the distilled Klein without conversion, at a slightly lower scale around 0.75.

How do I refer to several images in an edit?

Write image 1, image 2 and image 3 in the order you uploaded them, and say what each one brings. Put the LoRA's trigger or instruction first.

Per image. 2k costs more than 1k, and editing adds a charge for every source image. The number of LoRAs, the aspect ratio and the output format do not change the price. Failed requests are not charged.

FLUX 3 Klein Base LoRA API — Run Your Own LoRAs

black-forest-labs/flux-3-klein-base-lora/text-to-image

FLUX 3 Klein Base LoRA text to image runs Black Forest Labs' undistilled FLUX 3 Image Klein with up to three of your own LoRA files. Fifteen aspect ratios at 1k or 2k, JPEG, PNG or WebP.

From $0.042/image

Run model View API

Input

Text to image

Prompt433/5000

Aspect ratio

Resolutionaffects price

Style LoRAs1/3

Flux3 Image Klein Fennec Girl

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": "black-forest-labs/flux-3-klein-base-lora/text-to-image",
  "input": {
    "aspect_ratio": "3:4",
    "resolution": "1k",
    "loras": [
      {
        "path": "https://huggingface.co/comfyanonymous/flux3_image_klein_fennec_girl_lora/resolve/main/fennec_girl_flux3_klein_test_lora.safetensors",
        "scale": 0.6
      }
    ],
    "output_format": "jpeg",
    "seed": 401,
    "prompt": "fennec girl. Cinematic photograph, 50mm lens. An adult woman in her late twenties with large fennec fox ears and long auburn hair, wearing a cream off-shoulder knit sweater dress, sits on a reindeer hide beside a black wood stove in a timber cabin at blue dusk. Warm firelight on her face and bare shoulder, cold blue light behind her, she glances back at the camera with a sultry half smile. Soft film grain, realistic skin texture."
  }
}

Introduction

FLUX 3 Klein Base LoRA Text to image draws one image from a written prompt on Black Forest Labs' FLUX 3 Image Klein Base, the undistilled 5.6B checkpoint its makers recommend for LoRA training, with up to three of your own LoRA files loaded from direct links. Choose one of fifteen aspect ratios at 1k or 2k and JPEG, PNG or WebP. LoRA files are not reviewed, but without a matching LoRA the model often softens the strongest prompts; for those, use FLUX 3 Klein LoRA.

Choose a mode by starting material

What you have	Mode
a written description	FLUX 3 Klein Base LoRA Text to image
one to three images to change or combine	FLUX 3 Klein Base LoRA Image editing

Key features & highlights

Runs LoRAs on their own checkpoint

Black Forest Labs recommends training Klein LoRAs on Base, so a LoRA you trained runs here on exactly the model it learned from. In our tests a character LoRA trained on Base changed the face, ears and finish of the picture at 0.5 and more strongly at 1, while a LoRA made for FLUX.2 [klein] 9B loaded and changed nothing at all.

Exact A/B pairs

The same seed with the same inputs returned identical pixels in our tests. Run the prompt once without a LoRA and once with it, and every difference between the two pictures is the LoRA's doing, which makes judging a fresh training run much easier.

Up to three LoRAs

Stack up to three files, each with its own strength from 0 to 4, for example a character with a style beside it. LoRA links are downloaded per request; a file over 2 GB or a link to a model page fails the request without a charge.

Request parameters

Request parameters for FLUX 3 Klein Base LoRA · Text to Image
Parameter	Required	Description
`prompt`	Yes	Required, up to 5,000 characters. Put the LoRA's trigger phrase first, then a full caption: framing, subject, clothing, setting and light. at least 1 characters · at most 5000 characters
`aspect_ratio`	No	Fifteen shapes from 3:1 to 1:3; 1:1 by default. The shape does not change the price. `1:1` · `16:9` · `9:16` · `4:3` · `3:4` · `3:2` · `2:3` · `4:5` · `5:4` · `21:9` · `9:21` · `2:1` · `1:2` · `3:1` · `1:3` · Default: `1:1`
`resolution`	No	1k (default) is about one megapixel; 2k is about four megapixels, for example 1328 x 3120 for a tall picture. 2k costs more and takes about a minute. `1k` · `2k` · Default: `1k`
`loras`	No	Up to three {path, scale} entries trained for FLUX 3 Image Klein; scale 0 to 4, 1 by default. Use a public direct link to a .safetensors file up to 2 GB. Default: `[]` at most 3 items
`output_format`	No	JPEG (default), PNG or WebP. Does not change the price. `jpeg` · `png` · `webp` · Default: `jpeg`
`seed`	No	Optional, 0 to 2,147,483,647. In our tests the same seed and inputs returned identical pixels, so it isolates the effect of a LoRA. `≥ 0` · `≤ 2147483647`

Pricing

Billed per image. 2k costs more than 1k; the number of LoRAs, the aspect ratio and the output format do not change the price.

Pricing
Specification	Price
`1k`	$0.042 / image
`2k`	$0.15 / image

Use cases

Testing a fresh LoRA

Train on FLUX 3 Image Klein Base, then check the file here at 1k against the same seed without it, and at 2k for detail, without keeping a GPU of your own warm.

Characters and styles in series

Keep a character or a look consistent across fifteen shapes, from 3:1 banners to 1:3 phone screens, at one price per size tier.

Explore key use-case tutorials

Prompt guide & pro tips

Trigger first, then a full caption

Start with the trigger phrase from the LoRA's model card, then describe framing, subject, clothing, place and light in sentences. Klein is trained on long captions and this endpoint sends your prompt as written.

Start the scale around 0.5 to 1

A character LoRA at 1 also added a painted finish and a stray signature in our tests; at 0.5 it kept a photographic look. Move in steps of 0.2 and check faces and hands first.

Check the base model on the card

Only LoRAs for FLUX 3 Image Klein do anything here. If a run looks identical to the same seed without the LoRA, the file was most likely trained for another model.

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": "black-forest-labs/flux-3-klein-base-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",
    "resolution": "1k"
  }
}'

# 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?

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

LORA IMAGE API

FLUX 3 Klein Base LoRA API at a glance

The undistilled FLUX 3 Image Klein, the build LoRAs are trained on, with up to three of your own LoRA files.

LoRAs per call: 3; Up to three weight files per image, each with its own strength from 0 to 4.
Training checkpoint: Base; The undistilled Klein that Black Forest Labs recommends for LoRA training.
Aspect ratios: 15; From 3:1 to 1:3, including 21:9, 4:5 and 2:1.
Largest size tier: 4 MP; 2k is about four megapixels; 1k is about one megapixel.
Edit sources: 3; Refer to them as image 1, image 2 and image 3.
Softer without a LoRA: Often; Without a matching LoRA, prompts from the Strong level up mostly came back toned down.

Results

What you can make with FLUX 3 Klein Base LoRA

Character LoRA
A LoRA trained on Base, run on Base
A community character LoRA at 0.6 adds the ears and the face it was trained on, while the caption decides the cabin, the knitwear and the light.
Text to image with a character LoRA at 0.6, 3:4.
Edit LoRA
Outpainting that keeps the picture in place
Pad a portrait with flat gray, load an outpainting LoRA and use its instruction as the prompt: the gray becomes more of the room.
Outpainting LoRA at 1.5, 16:9.
Two sources
A cardigan from one photo, a person from another
Refer to the sources as image 1 and image 2; LoRAs can be added to the same request.
Edit combining a portrait and a cardigan photo.

More clips made with FLUX 3 Klein Base LoRA

Text to image with a character LoRA at 0.5, 4:5.
Outpainting LoRA at 1.5, 9:16.
Text to image without a LoRA, 2:3.

Content test

Is FLUX 3 Klein Base LoRA uncensored? What we tested

We ran each level on our live setup and checked every clip frame by frame. A pass means the content was actually on screen, not just that the request went through.

Content levels tested on FLUX 3 Klein Base LoRA, by mode
Level	Text to image
MildThe lightest prompts in the test	Not tested yet
ModerateWhere filtered models start to refuse	Not tested yet
StrongThe middle of the scale	SoftenedTested Oct 16, 2026
Very strongWhere most models start to hold back	Not tested yet
MaximumThe top of the scale	Not tested yet

Renders: on screen as written. Softened: generated, but milder than what was written. Refused: rejected. Not tested yet: we have not run that level on that mode.

How each mode behaves

All modes: Use LoRAs trained for FLUX 3 Image Klein. A FLUX.2 [klein] LoRA loaded without an error and changed nothing in our tests, and a camera-angle LoRA saved in diffusers format made no visible difference. Up to three LoRAs, scale 0 to 4; a link that needs a login, points to a web page or is larger than 2 GB fails the request.
All modes: The platform adds no content filter, but this undistilled variant often tones results down. Without a LoRA it drew the Mild and Moderate levels as asked in our tests, yet softened most prompts from the Strong level up, toned down a male subject, stopped the Very strong scene short of what was asked, and in editing handed the source back unchanged. If you need the highest levels, use the distilled FLUX 3 Klein instead.
All modes: It is slower than the distilled FLUX 3 Klein: Base runs about 50 sampling steps where the distilled build runs 16. In our tests a 1k image took about 14 to 45 seconds of model time, a 2k image about a minute, and a 2k edit close to 90 seconds.
Image editing: Editing takes one to three sources. With auto the result keeps the first image's shape at the chosen size tier. Each extra source adds to the price.
All modes: Every person you describe or upload must be an adult. You are responsible for the lawfulness of what you generate and publish, including the rights of anyone shown in a source image. Generated media is retained for 14 days and then deleted.

Use cases

What to build with the FLUX 3 Klein Base LoRA API

Checking a LoRA you trained
Run your file on the checkpoint it was trained on, at 1k or 2k, against the same seed without it, without hosting a GPU.
Edit LoRAs
Edit-trained LoRAs work on the editing endpoint, for example an outpainting LoRA that extends a picture into gray padding while keeping it in place.
Character and style LoRAs
Load a character or style LoRA with up to two more beside it and keep the strengths separate.
No filter on LoRAs
LoRA files are not reviewed. LoRAs for the stronger levels on FLUX 3 Klein were not yet published when we tested; the base model alone holds back from the Strong level up.

Prompts

FLUX 3 Klein Base LoRA prompt examples you can copy

Paste one into the playground as it is, then change one thing at a time.

Text to image
Trigger first, caption after
fennec girl. Cinematic photograph, 50mm lens. An adult woman in her late twenties with large fennec fox ears and long auburn hair, wearing a cream off-shoulder knit sweater dress, sits on a reindeer hide beside a black wood stove in a timber cabin at blue dusk. Warm firelight on her face and bare shoulder, cold blue light behind her, she glances back at the camera with a sultry half smile. Soft film grain, realistic skin texture.
Why it works. The LoRA's trigger opens the prompt, then a full caption sets the scene the LoRA should not decide.
Try this prompt
Image editing
The outpainting instruction
Outpaint the image: replace the solid gray areas with a seamless continuation of the scene, keeping the existing picture unchanged.
Why it works. Edit LoRAs expect their own instruction word for word; here it is the LoRA card's outpainting sentence.
This mode starts from your own image: add it in the playground before you run.
Try this prompt
Image editing
Two sources by number
Dress the woman from image 1 in the cream cable-knit cardigan from image 2, worn open and slipping off one shoulder over a black silk camisole. Keep her face, freckles and red hair. Seat her sideways on a pine bench in a timber sauna lounge with damp hair, glancing back over her shoulder at the camera. Cold blue daylight, warm amber stove glow, cinematic photograph, 50mm lens.
Why it works. Image 1 gives the person, image 2 the garment; the rest of the sentence says what stays and where she sits.
This mode starts from your own image: add it in the playground before you run.
Try this prompt

Pricing

FLUX 3 Klein Base LoRA API pricing

One image starts at $0.042 at 1K and $0.15 at 2K.

The playground shows the exact price for your settings before you run. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

See every price tier

Get started

How to use the FLUX 3 Klein Base LoRA API in three steps

Step 1
Use LoRAs made for FLUX 3 Klein
LoRAs trained on FLUX 3 Image Klein Base load here. LoRAs for FLUX.2 [klein] ran without an error and changed nothing in our tests: the picture was identical to one without them.
Step 2
Link the file itself
Use a public URL that downloads the .safetensors file. A link to a model page, or a file over 2 GB, fails the request without a charge.
Step 3
Compare against no LoRA
In our text-to-image tests a repeated seed returned identical pixels, so a pair with and without the LoRA shows exactly what it adds.

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.

Why SpicyAPI

Why run FLUX 3 Klein Base LoRA on SpicyAPI

No filter from us
We add no content filter of our own. What the model renders is what you get back.
Tested, not guessed
Content labels come from real generations we ran and checked, with the test date on the page.
Pay per result, in dollars
No subscription and no credits. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.
One key for every model
One API key and balance cover image, video, audio and chat models, so you can explore more models from the same account.

FAQ

FLUX 3 Klein Base LoRA API FAQ

About FLUX 3 Klein Base LoRA

FLUX 3 Klein Base LoRA runs the undistilled FLUX 3 Image Klein, the checkpoint LoRAs are trained on, behind up to three of your own LoRA files: text to image and editing with up to three images, and fifteen shapes at up to four megapixels.