Finetune

APIs for training new model.

Create Finetune

This endpoint will start the training of a LoRA model for Flux Pro and returns its associated unique trigger word.

SecurityAPIKeyHeader

Request

Request Body schema: multipart/form-data

image_files	Array of strings <binary> (Image Files) Images to be uploaded for training. For FluxFintune and FluxProFintune, providing a few images of the desired object/person/style will usually suffice. In case of FluxKontextFinetune, images should be provided in pairs. For each pair, the first image will be used as 'start' and the second image as 'end' for the training. Where the 'end' image contains the desired outcome of applying the edit described in 'prompt' to the 'start' image.
image_urls	Array of strings (Image URLs) Default: [] Alternatively you can provide URLs of the training image files. If you upload image files, this field will be ignored.
num_iterations	integer (Number of training iterations. Defaults to 0 for Auto. In Auto mode the actual values for ) Default: 0
option	string (Option) Default: "FluxFinetune" Training option to finetune based on your images. Available values are `FluxFinetune, FluxProFinetune, FluxKontextFinetune`.
finetune_type	string (Finetune Type) Default: "Auto" Override finetune type. Finetune type can be `Auto`, `Character`, `Product`, or `Style`. Choose `Auto` for automatic selection (default). This field is ignored when option is set to FluxKontextFinetune
lora_prefix	string (Lora Prefix) Default: "MYLORA" Prefix for the name of LoRA. Try to choose a unique string.
prompt	string (Prompt) Prompt describing the relationship between the start and end images in given image pairs when FluxKontextFinetune is used. It is ignored with other options. For instance, if in the 'start' image is a blue car but the 'end' image is a red car the prompt can be 'turn the car red.'
webhook_url	string (Webhook Url)

Responses

200

Trigger word.

400

Bad Request

402

Insufficient Credit

422

Validation Error

post/finetune

Request samples

curl
JavaScript
Python
Go
Node.js

Response samples

200
400
402
422

application/json

{"data": {"id": "ASSET_ID",
"customer_id": "CUSTOMER_ID",
"primary_user_id": "CUSTOMER_ID",
"args": { },
"results": ["result_url1",
"result_url2"
],
"thumbnails": [ ],
"status": "success",
"mime_type": "MIMETYPE",
"app_name": "FUNCTION_NAME",
"created_at": "2025-05-06 16:35:59.840508+00:00",
"updated_at": "2025-05-06 16:35:59.840508+00:00",
"history": [ ]
}
}

Finetune Flux Pro/Dev

This endpoint will report the status of training LoRA models for Flux Pro. If completed, it will also return sample outputs for the LoRA.