Introduction
Krea 3 Image editing changes or combines one to eight source images from a plain-language instruction. Krea built Krea 3 for edits first: say what should change and what should stay, and refer to the sources as image 1, image 2 and so on. You set the output shape; every result is about one megapixel.
Choose a mode by starting material
| What you have | Mode |
|---|---|
| a written description | Krea 3 Text to image |
| one to eight photos to change or combine | Krea 3 Image editing |
Key features & highlights
Combine people, clothes and places
Up to eight sources in one request. In our example the woman from image 1 wears the paisley dress from image 2 on a canyon deck, with her face and hair kept.
Name sources by number
"The person from image 2" picks the second upload: in our check with a woman as image 1 and a man as image 2, only the man appeared.
Identity holds through big changes
A woman in an office blazer became a 1970s portrait in a wood-panelled den with the same face, braids and framing.
Fashion-level content only
Requests and sources are screened. Outfit and scene changes went through; requests to remove clothing were refused and not charged.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | Say what to change, then what stays. Refer to sources as image 1, image 2 in upload order. at least 1 characters · at most 5000 characters |
image_urls | Yes | One to eight images, in the order the prompt names them. Each one adds the output price again. at least 1 items · at most 8 items |
aspect_ratio | No | Output shape: 1:1, 4:5, 3:4, 2:3, 9:16, 5:4, 4:3, 3:2, 16:9 or 21:9. Default 1:1; the source shape is not kept automatically. |
quality | No | high (default) favours detail; fast returns sooner. Same price. |
enable_thinking | No | Plans a longer prompt before drawing; about twice as slow in our tests. Off by default. Default: |
negative_prompt | No | What to keep out, up to 1,024 characters. Used only by the high-quality variant. at most 1024 characters |
output_format | No | PNG (default) or JPEG. No effect on price. |
seed | No | Optional, 0 to 2147483647. Reusing a seed brings you close to an earlier image but does not guarantee a match. |
Pricing
$0.072 / image
Billed per image: one output price plus the same amount for each source image. Quality variant, thinking and aspect ratio do not change the price. Refused requests are not charged.
Use cases
Virtual try-on and styling
A portrait plus a product photo of a garment becomes a styled shot of that person wearing it.
Restaging and period looks
Move a subject into a new location, era or light while keeping who they are.
Prompt guide & pro tips
Lead with the change
Start with what should change, then one sentence on what stays. For several sources, name each one: "the woman from image 1", "the dress from image 2".
Pick the output shape
The result uses the aspect ratio you choose rather than the source's shape. Match the ratio to the main photo to avoid unwanted cropping.
Mind the source count
Every source image is billed on top of the output, so send only the ones the prompt uses. State the age of people when you change their era or styling; looks can drift younger or older.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${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: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "krea/3/edit",
"input": {
"aspect_ratio": "1:1",
"image_urls": [
"https://example.com/reference.png"
],
"output_format": "png",
"prompt": "A ceramic teapot on a linen cloth, soft window light, muted palette",
"quality": "high"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.










