Which LoRAs work here?

Files trained for Krea 2, ideally on the Raw checkpoint as Krea recommends. Krea publishes its own style LoRAs, and Hugging Face and Civitai list many more. Pass the direct download link to the .safetensors file; both ComfyUI-style and Diffusers-style key names loaded in our tests. LoRAs made for FLUX, SDXL, Z-Image or Qwen-Image will not transfer.

Will it render NSFW images?

Not by default. The platform adds no filter, but Krea 2 Turbo itself tends to tone results down: in our tests most NSFW prompts came back softened. With a community unfiltered LoRA for Krea 2 loaded, the same prompts rendered as written, the Very strong level included.

Why did my LoRA request fail?

Usually the link needs a login (gated Hugging Face files answer with HTTP 401) or points to the repository page instead of the file. Use a public URL that downloads the .safetensors file itself. Failed requests are not charged.

My LoRA loaded but nothing changed. Why?

Either it was trained for a different base model, in which case it loads silently and has no effect, or it needs its trigger phrase. Check the LoRA page and put the trigger at the start of the prompt.

What strength should I use?

Start at 1 for a style LoRA and lower it toward 0.5 to season a look. The field accepts up to 4, but in our test that value broke the image into flat shapes.

Which Krea 2 checkpoint does this run?

Krea 2 Turbo, the eight-step distilled checkpoint Krea released as open weights on 22 June 2026. Krea's own advice is to train LoRAs on the undistilled Raw checkpoint and run them on Turbo, which is the pairing this family is built for.

Krea 2 Turbo LoRA API: NSFW Images With Your Own LoRAs

krea/2-turbo-lora/text-to-image

Text-to-image on Krea 2 Turbo with up to three of your own LoRAs per request, each with its own scale. Plain-sentence prompts, fifteen aspect ratios, 1k or native 2k, and no platform filter on prompts, weights or results.

From $0.022/image

Run model View API

Input

Text to image

Prompt592/5000

Aspect ratio

Resolutionaffects price

Style LoRAs0/3

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": "krea/2-turbo-lora/text-to-image",
  "input": {
    "aspect_ratio": "4:5",
    "resolution": "1k",
    "seed": 11,
    "prompt": "Cinematic portrait inside an old tack room at golden hour. A Latina woman in her early thirties with dark curls pinned loosely at the nape sits on a wooden saddle rack, a cream silk blouse unbuttoned at the collar and slipping off one shoulder, sleeves rolled, tan breeches and tall oxblood boots. She polishes a brass stirrup with a soft cloth and gives the camera a slow, smouldering glance. Rows of oiled saddles, bridles on brass hooks and a dusty window throwing a low amber shaft of light across the leather. 85mm lens, shallow depth of field, warm Kodak Portra colour, fine film grain."
  }
}

Introduction

Krea 2 Turbo LoRA Text to image draws one image from a written prompt with up to three of your own LoRA files loaded. It runs on Krea 2 Turbo, the eight-step distilled checkpoint of Krea's open-weight 12-billion-parameter image model, which Krea released alongside an undistilled Raw checkpoint meant for training LoRAs. Pick any of fifteen aspect ratios and one of two resolution tiers. The platform adds no content filter to the prompt, the weights or the image.

Choose a mode by starting material

What you have	Mode
a written description	Krea 2 Turbo LoRA Text to image
a picture to restyle	Krea 2 Turbo LoRA Image to image

Key features & highlights

Krea's own LoRAs and community ones

Krea publishes style LoRAs for this model, such as a soft Art Deco watercolour, a monochrome ink wash and an ethereal motion blur, and the community has trained many more on the Raw checkpoint. Pass the direct file link and a scale; both ComfyUI-style and Diffusers-style files loaded in our tests.

Three at once

Up to three LoRAs per request, each with its own scale. In our test a watercolour style at 0.8, a second LoRA at 0.8 and an ink wash at 0.3 loaded together and the watercolour look still led.

Eight steps, up to 2k

Turbo is distilled to eight sampling steps. Most of our test images took 5 to 20 seconds on 1k; the native 2k tier takes roughly two to three times as long and returns about four megapixels.

Toned down by default, open with a LoRA

Krea 2 Turbo tends to soften unfiltered prompts: most of our test prompts came back toned down, and nothing was refused. With a community unfiltered LoRA for Krea 2 loaded, the same prompts rendered as written.

Request parameters

Request parameters for Krea 2 Turbo LoRA · Text to Image
Parameter	Required	Description
`prompt`	Yes	Put the LoRA's trigger phrase first, then describe the picture in plain sentences: subject, clothing, place, light, lens. Up to 5,000 characters; there is no negative prompt. at least 1 characters · at most 5000 characters
`aspect_ratio`	No	Fifteen options from 3:1 to 1:3; 1:1 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` · `2:1` · `1:2` · `3:1` · `1:3` · `5:4` · `4:5` · Default: `1:1`
`resolution`	No	1k is about one megapixel and is the default. 2k is about four megapixels, generated natively, and costs more. `1k` · `2k` · Default: `1k`
`loras`	No	Up to three {path, scale} entries trained for Krea 2; scale 0 to 4, 1 is neutral and 0.5 to 1.2 is the useful range. Use a public direct link to the .safetensors file. Default: `[]` at most 3 items
`seed`	No	Keeps a set of images comparable while you change one LoRA scale at a time. It does not guarantee the exact same pixels on every run. `≥ 0` · `≤ 2147483647`

Pricing

Billed per image by resolution tier. The aspect ratio does not change the price, and loading LoRAs adds no separate charge.

Pricing
Specification	Price
`1k`	$0.0216 / image
`2k`	$0.036 / image

Use cases

Style batches

A style LoRA plus a list of prompts becomes a consistent set of illustrations, posters or thumbnails. Fix the seed per layout and change only the subject line.

Testing your own LoRA

Train on Krea 2 Raw, then check the LoRA here at Turbo speed across several prompts and scales before you build on it.

Unfiltered stills with community weights

Load the LoRA your audience expects and write the image directly. There is no content flag and no second review of the result.

Explore key use-case tutorials

Prompt guide & pro tips

Trigger phrase first, then sentences

Krea's style LoRAs use plain trigger phrases such as "Art Deco watercolor style". Put it at the very start, then describe subject, clothing, place, light and lens in full sentences; Krea 2 reads natural language well.

Compare against no LoRA

Run the same prompt and seed without the LoRA. A style trigger alone already nudges the base model toward that medium, so judge the LoRA by the difference, not by whether the style appears.

Stay between 0.5 and 1.2

The scale field accepts 0 to 4, but at 4 our test image collapsed into flat geometric shapes. 0.5 already reads clearly as the style.

Use direct, public links

The link must download the .safetensors file itself. A repository page or a file that needs a login fails the request, which is then not charged.

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": "krea/2-turbo-lora/text-to-image",
  "input": {
    "aspect_ratio": "1:1",
    "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

Krea 2 Turbo LoRA API at a glance

Krea's open 12B image model with your own LoRAs: eight steps, up to 2k, nothing filtered on our side.

LoRAs per call: 3; Up to three weight files per image, each with its own strength. ComfyUI and Diffusers key names both load.
Strength range: 0 – 4; 1 is neutral. In our tests 0.5 already showed a style clearly and 4 flattened the image into shapes.
Sampling steps: 8; Krea's distilled Turbo checkpoint. Most of our test images took 5 to 20 seconds.
Aspect ratios: 15; From 3:1 to 1:3 on both tiers. The ratio never changes the price.
Top resolution: 2k; About one megapixel on 1k, about four megapixels on 2k, generated natively rather than upscaled.
Platform filter: None; Nothing is refused or blurred. The base model tones results down on its own; an unfiltered LoRA changes that.

Results

What you can make with Krea 2 Turbo LoRA

Style
Krea's own watercolour LoRA
"Art Deco watercolor style" at the start of the prompt and Krea's softwatercolor LoRA at full strength: flat washes, ink outlines and Deco geometry in the carriage panels, on a scene the base model would otherwise render as a photograph.
Text-to-image with Krea's watercolour LoRA at 1, 16:9 at 2k.
Ink
Monochrome ink wash from one LoRA
The darkbrush LoRA with its trigger phrase turns a foggy paddock into black ink on white paper: dry-brush fence rails, a wet-on-wet sky and fog left as bare paper.
Text-to-image with Krea's ink wash LoRA at 1, 3:2 at 2k.
Restyle
A photograph in a custom style
Image-to-image at strength 0.6 with the watercolour LoRA: the woman, her tweed jacket, the bridle and the brass lanterns stay where they were in the photo, redrawn as an illustration.
Image-to-image with the watercolour LoRA at strength 0.6.

More clips made with Krea 2 Turbo LoRA

Text-to-image without a LoRA, 4:5 at 2k.
Image-to-image without a LoRA at strength 0.6, 2k.
Image-to-image with the motion blur LoRA at strength 0.6.
Image-to-image with Krea's stippling LoRA at strength 0.55.
Text-to-image without a LoRA, 2:3 at 2k.
Text-to-image with Krea's motion blur LoRA at 1, 2:1 at 2k.

Content test

Is Krea 2 Turbo 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 Krea 2 Turbo LoRA, by mode
Level	Image editing	Text to image
MildThe lightest prompts in the test	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
ModerateWhere filtered models start to refuse	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
StrongThe middle of the scale	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
Very strongWhere most models start to hold back	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
MaximumThe top of the scale	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026

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: Out of the box Krea 2 Turbo tones results down: in our tests most NSFW requests came back softened, and Very strong and Maximum scenes were turned into tamer versions. Nothing is refused or blurred, and the image is charged as normal. An unfiltered LoRA trained for Krea 2 changes that. Weights must be trained for Krea 2; a LoRA for another base model loads without an error and changes nothing. The file is fetched when the request runs, so a large or new LoRA adds time, and a link that needs a login or points to a web page can fail the request. Some style LoRAs leave marks of their own: Krea's ink wash LoRA often painted a small signature-like squiggle in a corner in our tests, so reroll the seed if one appears. There is no negative prompt and no guidance or step control; one JPEG per request.
Image editing: The result keeps the source's aspect ratio and the whole frame is redrawn; there is no mask. Strength acts early: at 0.3 details already change, around 0.5 clothing and lighting follow the prompt, and from 0.8 the scene itself is replaced.
All modes: You are responsible for the lawfulness of what you publish. Do not generate material involving minors, non-consensual intimate imagery, or deceptive impersonation of a real person. That is a boundary on use, not a claim that each request is screened.

Use cases

What to build with the Krea 2 Turbo LoRA API

A house style for a whole set
Load a style LoRA and run a list of prompts. Watercolour, ink wash or a soft motion-blur look stays consistent across every image, which prompting the base model alone does not promise.
Your own trained LoRA, at Turbo speed
Train a character or product LoRA on Krea 2 Raw with the usual tools and pass its link here. Eight-step sampling keeps iterations quick.
Photos restyled in one look
Image-to-image with a style LoRA at strength 0.55 to 0.6 keeps the composition of a photo and redraws it in the LoRA style.
Unfiltered work with community LoRAs
Unfiltered LoRAs load without review and nothing is filtered on the way out. The base model tones results down by default; with an unfiltered LoRA loaded, our test prompts rendered as written at every level up to Maximum.

Prompts

Krea 2 Turbo LoRA prompt examples you can copy

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

Text to image
Tack room, leather and brass
Cinematic portrait inside an old tack room at golden hour. A Latina woman in her early thirties with dark curls pinned loosely at the nape sits on a wooden saddle rack, a cream silk blouse unbuttoned at the collar and slipping off one shoulder, sleeves rolled, tan breeches and tall oxblood boots. She polishes a brass stirrup with a soft cloth and gives the camera a slow, smouldering glance. Rows of oiled saddles, bridles on brass hooks and a dusty window throwing a low amber shaft of light across the leather. 85mm lens, shallow depth of field, warm Kodak Portra colour, fine film grain.
Why it works. No LoRA: the base model on its own at 1k. Lens, light and film stock in the prompt carry the look.
Try this prompt
Image editing
Stable yard as a watercolour
Art Deco watercolor style. A Middle Eastern woman in her thirties with long dark wavy hair, in a brown herringbone tweed jacket worn open over a white shirt unbuttoned at the throat, holds a leather bridle with brass buckles in a cobbled stable yard, brass carriage lanterns and green stable doors behind her, soft morning light.
Why it works. A photo in, the watercolour LoRA at 1 and strength 0.6. The pose, bridle and lanterns are kept; the medium changes.
This mode starts from your own image: add it in the playground before you run.
Try this prompt
Text to image
Art deco carriage house
Art Deco watercolor style. Inside an old carriage house at dusk, a lacquered black carriage with polished brass lamps. An East Asian woman in her thirties with a sleek black bob, in a bias-cut champagne satin slip dress with a navy riding jacket draped over her bare shoulders, leans against the tall carriage wheel. A South Asian man in his thirties with a neat moustache, in a charcoal waistcoat and a white shirt with the sleeves rolled, stands close with one hand on the brass lamp, their faces inches apart. Lantern glow, long shadows, straw on the cobbles.
Why it works. Krea's softwatercolor LoRA at 1 with its trigger phrase first. 16:9 at 2k returns 2720 by 1536 pixels.
Try this prompt

Pricing

Krea 2 Turbo LoRA API pricing

One image starts at $0.022 at 1K and $0.036 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 Krea 2 Turbo LoRA API in three steps

Step 1
Put the trigger phrase first
Krea's own style LoRAs each have a trigger phrase, such as "Art Deco watercolor style" or "monochrome ink wash style". Start the prompt with it, then describe the picture in plain sentences.
Step 2
Use LoRAs trained on Krea 2 Raw
Krea's recommendation is to train on the Raw checkpoint and run on Turbo, and that is what this endpoint expects. A LoRA trained for another base model loads without an error and changes nothing that matters.
Step 3
Start at strength 1
Scale runs from 0 to 4. In our tests 0.5 was already a clear style and 1 was the full look; at 4 the picture fell apart into flat geometric shapes.

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 Krea 2 Turbo 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

Krea 2 Turbo LoRA API FAQ

About Krea 2 Turbo LoRA

Krea 2 Turbo LoRA puts Krea's open eight-step image model behind your own LoRAs: text-to-image in fifteen aspect ratios at 1k or native 2k and image-to-image with a strength control, with up to three weight files per request. Style, character and unfiltered LoRAs load without review, and the platform filters nothing.

All Krea models5

Related models and next tasks

LoRA guide

Which LoRAs load on Krea 2 Turbo LoRA

Workbench

Input

Output

Introduction

Choose a mode by starting material

Key features & highlights

Krea's own LoRAs and community ones

Three at once

Eight steps, up to 2k

Toned down by default, open with a LoRA

Request parameters

Pricing

Use cases

Style batches

Testing your own LoRA

Unfiltered stills with community weights

Prompt guide & pro tips

Trigger phrase first, then sentences

Compare against no LoRA

Stay between 0.5 and 1.2

Use direct, public links

API quick start

Krea's own watercolour LoRA

Monochrome ink wash from one LoRA

A photograph in a custom style

More clips made with Krea 2 Turbo LoRA

How each mode behaves

A house style for a whole set

Your own trained LoRA, at Turbo speed

Photos restyled in one look

Unfiltered work with community LoRAs

Tack room, leather and brass

Stable yard as a watercolour

Art deco carriage house

Put the trigger phrase first

Use LoRAs trained on Krea 2 Raw

Start at strength 1

No filter from us

Tested, not guessed

Pay per result, in dollars

One key for every model

Which LoRAs work here?+

Will it render NSFW images?+

Why did my LoRA request fail?+

My LoRA loaded but nothing changed. Why?+

What strength should I use?+

Which Krea 2 checkpoint does this run?+

About Krea 2 Turbo LoRA

All Krea models5

Related models and next tasks

LoRA guide