LoRA generation

Generate images that follow your LoRA (custom-trained style or subject). Pass the LoRA identifiers you use in TheFluxTrain together with a prompt.

URL: POST https://api.thefluxtrain.com/api/v1/image-generate/lora

Authentication: x-api-key

Request body

Field	Type	Required	Default	Description
`prompt`	string	Yes	—	Text prompt
`loras`	string	Yes	—	LoRA identifiers (e.g. Hugging Face repo names or URLs)
`trainer_model`	string	Yes	—	`flux1`, `flux2`, or `qwen_image`
`lora_scales`	number	No	`[1.0]`	Scale per LoRA (0.0–2.0)
`aspect_ratio`	string	No	`1:1`	Output aspect ratio
`prompt_strength`	number	No	`0.8`	Prompt strength
`num_outputs`	integer	No	`1`	Number of images
`num_inference_steps`	integer	No	`28`	Inference steps
`guidance_scale`	number	No	`3.5`	Guidance scale
`output_format`	string	No	`png`	Output format
`output_quality`	integer	No	`80`	Output quality
`image`	string	No	—	Input image URL or data URI (image-to-image)
`mask`	string	No	—	Mask URL or data URI
`seed`	integer	No	—	Seed for reproducibility
`use_oyecartoon`	boolean	No	`false`	Cartoon-style path when supported
`should_paste_reference`	boolean	No	`false`	Paste reference image
`strength`	number	No	`0.9`	Image-to-image strength
`folder_id`	string	No	—	Folder UUID in your account
`project_id`	string	No	—	Optional project ID
`training_id`	string	No	—	Optional training ID
`message_id`	string	No	—	Your own tracking ID

Example

{
  "prompt": "ohwx person in a cafe, natural light",
  "loras": ["your-org/your-lora"],
  "trainer_model": "flux2",
  "lora_scales": [1.0],
  "aspect_ratio": "3:4",
  "folder_id": "your-folder-uuid"
}

Response

Same response format as Image generation. Key data fields:

Field	Type	Description
`id`	string	Job ID
`url`	string	Output URL when complete
`status`	string	`SUCCESS`, `PENDING`, etc.
`media_type`	string	`IMAGE`
`meta`	object	Provider metadata
`folder_id`	string \| null	Folder UUID
`created_at`	string \| null	ISO 8601 timestamp

Poll Jobs when status is PENDING.

Documentation

Request body

Example

Response