Skip to content

Krea 2 Edit API

Krea 2 Identity Edit behind one REST call: send a photo of a person and an instruction, get the same person in the scene, outfit, pose and lighting you described. The reference_strength parameter decides how closely the result sticks to the photo.

Price
$0.03 / image
Model id
krea-2-edit
Reference images
1
Sizes
7 presets

1024x1024, 832x1216, 1216x832, 896x1152, 1152x896, 1536x1024, 1024x1536

What it is good at

krea-2-edit runs the community Krea 2 Identity Edit LoRA on a Krea 2 diffusion transformer with the dual conditioning it was trained with: the photo is injected as in-context image tokens and also read by the Qwen3-VL text encoder alongside your instruction, so the model understands both what the person looks like and what you asked for.

It is the model for people. Where a general edit model drifts away from a face after a few words of instruction, this one holds identity while you change everything around it: scene, pose, camera angle, season, outfit, hairstyle, lighting. For edits that are not about a person — recolouring objects, swapping backgrounds on products, combining several references — qwen-image-edit is the better fit.

  • Restaging: the same person in a new place, pose or era
  • Wardrobe and styling changes described in plain language
  • Relighting: flash, golden hour, neon, candlelight, studio
  • Head and face swaps, try-on and outpainting from the v1.2 training set
  • Bring your own Krea 2 LoRAs by URL (up to 3) behind the edit LoRA

Edit a photo

A minimal request needs a model, one image and an instruction. The output size defaults to the supported size closest to the photo’s aspect ratio.

curl
curl https://api.fastgencloud.com/v1/images/edits \
  -H "Authorization: Bearer $FASTGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "krea-2-edit",
    "image": "https://example.com/portrait.jpg",
    "prompt": "Put her on a sunny beach at golden hour in a white linen dress, laughing at the camera",
    "reference_strength": 2
  }'

How strongly to hold the reference

reference_strength multiplies how much the generated image attends to the photo. At 1 the reference is neutral and the instruction restages freely; the default of 2 holds the face and build while an explicit instruction changes outfit, pose and scene; at 4 the photo’s clothing and props carry over as well unless the instruction says otherwise; above 6 the output starts to copy the photo wholesale. Start at 2 for restaging, go to 4 when the person should look exactly as photographed, and toward 6 for lighting or background-only edits.

Prompting matters more than the dial: describe the whole target image (scene, outfit, pose, light, camera) rather than only the change, and say what should stay the same.

1
neutral: identity flows, layout, pose and outfit follow the instruction
2 (default)
face and build held; outfit, pose and scene change when the instruction names them
3–4
as photographed: clothing and props carry over unless told otherwise
6–8
near-copy of the photo: use for lighting or background-only edits

Parameters

model
krea-2-edit
image
one reference photo (URL or data URI); images[] with more than one entry is rejected
prompt
the edit instruction, 1–2000 characters, passed verbatim
reference_strength
0.5–8, default 2
size
1024x1024, 832x1216, 1216x832, 896x1152, 1152x896, 1536x1024, 1024x1536 (default: closest to the photo’s aspect ratio)
seed
optional integer for reproducible output
steps / guidance
default 10 steps; steps 4–20, guidance fixed at 1 (negative prompts have no effect)
loras
optional [{ url, strength }] — public Krea 2 LoRA files, up to 3
webhook_url
optional signed callback on completion

Unfiltered and uncensored

This endpoint applies no content filter. Your prompt, negative prompt, reference images and LoRAs reach the model verbatim, nothing is rewritten or injected, and outputs are not scanned or blocked. What your product allows is your call.

The one automatic rule is the minor-safety floor: a prompt that pairs a minor or age reference with sexual content is refused for that request only. Nothing else is filtered and the account is never touched by moderation.

How the API works

Every generation is a job. The POST returns 202 with a job id immediately, so your request thread is never blocked on a GPU. You then either poll GET /v1/jobs/{id} or register a webhook_url and receive a signed callback when the job finishes.

Send an Idempotency-Key header and a retried request replays the original job instead of charging you twice. Your wallet is debited when the job is submitted and refunded automatically if it fails.

poll the job
curl https://api.fastgencloud.com/v1/jobs/$JOB_ID \
  -H "Authorization: Bearer $FASTGEN_API_KEY"

Pricing and billing

Pricing is prepaid and usage-based: you top the wallet up with a card and each job debits it. There is no subscription, no per-seat cost, and no monthly minimum.

Images above 1 megapixel are billed pro rata by pixel count — a 1.5 MP image costs 1.5x the listed rate. Failed jobs are refunded in full, automatically.

Frequently asked questions

Is this the official Krea AI edit API?
No. This is an independent host of the open-weight Krea 2 model with the community Krea 2 Identity Edit LoRA (v1.2) and its ComfyUI node pack, under the Krea 2 Community License. Krea’s own products are separate.
How is it different from qwen-image-edit?
qwen-image-edit is a general instruction editor that takes up to three references and excels at objects, backgrounds and compositing. krea-2-edit takes one photo of a person and is tuned to keep that person recognisable while the rest of the image changes, with a dial for how tightly to hold them.
Can I keep the person and put them into another photo of a scene?
Not yet on this model id: it takes a single reference, the person. Describe the target scene in the instruction instead, or train a character with the characters API and generate the scene on the Krea 2 lineup.
Are my prompts or images modified before they reach the model?
No. The instruction is encoded exactly as you send it, together with your photo, and there are no injected negatives or added style tags.
How long are generated images stored?
Output URLs live for up to 7 days, then the files are deleted. Download anything you want to keep and store it yourself.

Also searched as

  • Krea 2 Edit API
  • Krea 2 identity edit API
  • identity-preserving image editing API
  • restage a person from one photo API
  • put a person in a different scene API
  • change outfit from photo API
  • Krea 2 image edit API
  • reference strength image editing
  • Krea 2 Identity Edit LoRA hosted
  • instruction-based photo editing API

Related

Start generating

Create an account, add a prepaid balance, and call the API with a key from the console. No subscription, no minimum, no per-seat pricing.

Get an API key