AI-powered fashion image generation — virtual try-on, model creation, background editing, and more

The Fashion API gives you access to 10 AI-powered fashion image generation endpoints through a single Concurred API key. Powered by FASHN, it supports virtual try-on, model creation, background manipulation, image editing, and video generation.

Async Pattern

Unlike the Chat API (which streams text), the Fashion API uses an async submit-and-poll pattern. You submit a job, get a prediction ID, then poll for results. We also provide a convenience subscribe endpoint that handles polling for you via SSE.

Available Models

Model Name	Description	Required Inputs	Credits	Speed
`tryon-v1.6`	Virtual try-on at 864×1296 resolution	`model_image`, `garment_image`	3	5-17s
`product-to-model`	Flat-lay/product image → on-model visual	`product_image`	3	~12s
`face-to-model`	Face → try-on ready upper-body avatar	`face_image`	3	~12s
`model-create`	Generate fashion models from text prompt	`prompt`	3	~12s
`model-variation`	Create variations of existing model images	`model_image`	3	8-10s
`model-swap`	Replace model identity, keep clothing/pose	`model_image`	3	10-12s
`reframe`	Smart aspect ratio adjustment with generative fill	`image`	2	~12s
`background-change`	Replace background with AI-generated scene	`image`, `prompt`	2	~12s
`background-remove`	Remove background → transparent PNG cutout	`image`	2	1-3s
`image-to-video`	Single image → fashion video clip	`image`	5	1-3min

Endpoints Overview

Method	Endpoint	Description
`POST`	`/api/v1/fashion/run`	Submit a generation job (returns prediction ID)
`GET`	`/api/v1/fashion/status/:id`	Poll job status and retrieve results
`POST`	`/api/v1/fashion/subscribe`	Submit and stream status updates via SSE

Quick Start

# 1. Submit a virtual try-on job
curl -X POST https://agent-heavy.vercel.app/api/v1/fashion/run \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "tryon-v1.6",
    "inputs": {
      "model_image": "https://example.com/person.jpg",
      "garment_image": "https://example.com/shirt.jpg"
    }
  }'
 
# Response: { "data": { "prediction_id": "abc-123", "status": "starting" } }
 
# 2. Poll for results
curl https://agent-heavy.vercel.app/api/v1/fashion/status/abc-123 \
  -H "Authorization: Bearer YOUR_API_KEY"
 
# Response when complete:
# { "data": { "status": "completed", "output": ["https://cdn.fashn.ai/.../output_0.png"] } }

Submit Job

POST /api/v1/fashion/run

Submits a fashion generation job. Returns immediately with a prediction ID.

Request Body

{
  "model_name": "tryon-v1.6",
  "inputs": {
    "model_image": "https://example.com/person.jpg",
    "garment_image": "https://example.com/garment.jpg",
    "category": "auto",
    "mode": "balanced"
  }
}

Parameter	Type	Required	Description
`model_name`	string	Yes	One of the 10 model names listed above
`inputs`	object	Yes	Model-specific input parameters (see each model's section below)

Response

{
  "success": true,
  "data": {
    "prediction_id": "123a87r9-4129-4bb3-be18-9c9fb5bd7fc1-u1",
    "status": "starting",
    "model_name": "tryon-v1.6",
    "created_at": 1709500000000
  },
  "meta": {
    "requestId": "uuid",
    "timestamp": "2026-03-04T00:00:00.000Z"
  }
}

Poll Status

GET /api/v1/fashion/status/:id

Check the current status of a prediction and retrieve results when complete.

Status Values

Status	Description
`starting`	Job is being initialized
`in_queue`	Waiting to be processed
`processing`	Model is actively generating
`completed`	Done — output URLs available
`failed`	Failed — check `error` field

Response (Completed)

{
  "success": true,
  "data": {
    "prediction_id": "123a87r9-4129-4bb3-be18-9c9fb5bd7fc1-u1",
    "status": "completed",
    "output": [
      "https://cdn.fashn.ai/123a87r9/output_0.png"
    ],
    "error": null
  }
}

Response (Failed)

{
  "success": true,
  "data": {
    "prediction_id": "123a87r9-4129-4bb3-be18-9c9fb5bd7fc1-u1",
    "status": "failed",
    "output": null,
    "error": {
      "name": "ImageLoadError",
      "message": "Error loading model image: Invalid URL format"
    }
  }
}

Output Availability

CDN URLs are available for 72 hours after completion. If you use return_base64: true, base64 outputs are available for 60 minutes. Download or cache results promptly.

POST /api/v1/fashion/subscribe

Convenience endpoint that submits a job and streams real-time status updates via Server-Sent Events. No manual polling needed.

Request Body

Same as /api/v1/fashion/run.

SSE Events

event: fashion
data: {"type":"job_submitted","prediction_id":"abc-123","model_name":"tryon-v1.6"}

event: fashion
data: {"type":"status_update","prediction_id":"abc-123","status":"processing"}

event: fashion
data: {"type":"completed","prediction_id":"abc-123","status":"completed","output":["https://cdn.fashn.ai/.../output_0.png"],"error":null}

data: [DONE]

Event Type	Description
`job_submitted`	Job accepted, includes prediction ID
`status_update`	Status transitioned (e.g. `in_queue` → `processing`)
`completed`	Job finished, includes `output` array
`failed`	Job failed, includes `error` object
`error`	Unexpected error during polling

JavaScript Example

const response = await fetch(`${BASE_URL}/api/v1/fashion/subscribe`, {
  method: 'POST',
  headers: HEADERS,
  body: JSON.stringify({
    model_name: 'background-remove',
    inputs: { image: 'https://example.com/photo.jpg' },
  }),
});
 
const reader = response.body.getReader();
const decoder = new TextDecoder();
 
while (true) {
  const { done, value } = await reader.read();
  if (done) break;
 
  const text = decoder.decode(value);
  for (const line of text.split('\n')) {
    if (line.startsWith('data: ') && line !== 'data: [DONE]') {
      const event = JSON.parse(line.slice(6));
      console.log(event.type, event);
 
      if (event.type === 'completed') {
        console.log('Output URLs:', event.output);
      }
    }
  }
}

Timeout

The subscribe endpoint has a 2-minute timeout. For long-running jobs like image-to-video (1-3 minutes), use the async /run + /status pattern instead.

Model Reference

Virtual Try-On (`tryon-v1.6`)

Try a garment on a person. Outputs at 864×1296 resolution.

Parameter	Type	Required	Description
`model_image`	string	Yes	URL or base64 of the person image
`garment_image`	string	Yes	URL or base64 of the garment image
`category`	string	No	`auto`, `tops`, `bottoms`, `one-pieces` (default: `auto`)
`mode`	string	No	`performance` (5s), `balanced` (8s), `quality` (12-17s)
`garment_photo_type`	string	No	`auto`, `flat-lay`, `model`
`moderation_level`	string	No	`conservative`, `permissive`, `none`
`num_samples`	number	No	1-4 images per request (default: 1)
`segmentation_free`	boolean	No	Disable clothing segmentation (default: true)
`seed`	number	No	0 to 2³²-1 for reproducibility
`output_format`	string	No	`png` or `jpeg` (default: `png`)
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "tryon-v1.6",
  "inputs": {
    "model_image": "https://example.com/person.jpg",
    "garment_image": "https://example.com/tshirt.jpg",
    "category": "tops",
    "mode": "quality"
  }
}

Product to Model (`product-to-model`)

Transform flat-lay or product-only images into realistic on-model visuals. Supports two modes: generate a new person (default) or add product to an existing model image.

Parameter	Type	Required	Description
`product_image`	string	Yes	URL or base64 of the product
`model_image`	string	No	Person image (enables try-on mode)
`image_prompt`	string	No	Inspiration image for pose/environment
`prompt`	string	No	Styling instructions (e.g. "athletic man, studio background")
`aspect_ratio`	string	No	`1:1`, `2:3`, `3:4`, `4:5`, `5:4`, `4:3`, `3:2`, `16:9`, `9:16`
`resolution`	string	No	`1k` (precise) or `4k` (creative)
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg`
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "product-to-model",
  "inputs": {
    "product_image": "https://example.com/jacket-flatlay.jpg",
    "prompt": "athletic woman, outdoor setting"
  }
}

Face to Model (`face-to-model`)

Transform a face/selfie into a try-on ready upper-body avatar.

Parameter	Type	Required	Description
`face_image`	string	Yes	URL or base64 of the face image
`prompt`	string	No	Body guidance (e.g. "athletic build", "curvy figure")
`aspect_ratio`	string	No	`1:1`, `4:5`, `3:4`, `2:3`, `9:16` (default: `2:3`)
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg` (default: `jpeg`)
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "face-to-model",
  "inputs": {
    "face_image": "https://example.com/selfie.jpg",
    "prompt": "athletic build",
    "aspect_ratio": "3:4"
  }
}

Model Create (`model-create`)

Generate realistic fashion models from text prompts, optionally guided by a reference image.

Parameter	Type	Required	Description
`prompt`	string	Yes	Description of desired model, clothing, pose, scene
`image_reference`	string	No	Reference image for composition guidance
`reference_type`	string	No	`pose` or `silhouette` (when `image_reference` is provided)
`aspect_ratio`	string	No	`1:1`, `2:3`, `3:4`, `4:5`, `5:4`, `4:3`, `3:2`, `16:9`, `9:16`
`lora_url`	string	No	FLUX-compatible LoRA weights URL (.safetensors, max 256MB)
`disable_prompt_enhancement`	boolean	No	Use prompt exactly as-is
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg`
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "model-create",
  "inputs": {
    "prompt": "Young woman in a red summer dress, standing on a beach at sunset",
    "aspect_ratio": "2:3"
  }
}

Model Variation (`model-variation`)

Create controlled variations from an existing model image.

Parameter	Type	Required	Description
`model_image`	string	Yes	Source fashion model image
`variation_strength`	string	No	`subtle` or `strong` (default: `subtle`)
`lora_url`	string	No	FLUX-compatible LoRA weights URL
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg`
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "model-variation",
  "inputs": {
    "model_image": "https://example.com/model.jpg",
    "variation_strength": "strong"
  }
}

Model Swap (`model-swap`)

Replace the model's identity (face, skin tone, hair) while keeping clothing and pose intact.

Parameter	Type	Required	Description
`model_image`	string	Yes	Source image with clothing/pose to preserve
`prompt`	string	No	Desired identity description (e.g. "Asian woman with short hair")
`background_change`	boolean	No	Also change background based on prompt
`disable_prompt_enhancement`	boolean	No	Use prompt exactly as-is
`lora_url`	string	No	FLUX-compatible LoRA weights URL
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg`
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "model-swap",
  "inputs": {
    "model_image": "https://example.com/model-in-outfit.jpg",
    "prompt": "Middle-aged man with dark hair and beard"
  }
}

Reframe (`reframe`)

Adjust image aspect ratios using AI-powered generative fill. Two modes: directional extension or aspect ratio adjustment.

Parameter	Type	Required	Description
`image`	string	Yes	Source image to reframe
`mode`	string	No	`direction` or `aspect_ratio`
`target_aspect_ratio`	string	No	Target ratio (for `aspect_ratio` mode): `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`
`target_direction`	string	No	Expand direction (for `direction` mode): `both`, `down`, `up`
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg`
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "reframe",
  "inputs": {
    "image": "https://example.com/portrait.jpg",
    "mode": "aspect_ratio",
    "target_aspect_ratio": "16:9"
  }
}

Background Change (`background-change`)

Replace the background with an AI-generated scene while preserving the foreground subject.

Parameter	Type	Required	Description
`image`	string	Yes	Source image
`prompt`	string	Yes	Description of new background (e.g. "beach sunset", "modern office")
`disable_prompt_enhancement`	boolean	No	Use prompt exactly as-is
`seed`	number	No	0 to 2³²-1
`output_format`	string	No	`png` or `jpeg`
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "background-change",
  "inputs": {
    "image": "https://example.com/model-photo.jpg",
    "prompt": "Luxury boutique interior with warm lighting"
  }
}

Background Remove (`background-remove`)

Remove the background and create a transparent PNG cutout.

Parameter	Type	Required	Description
`image`	string	Yes	Source image
`return_base64`	boolean	No	Return base64 instead of CDN URL

{
  "model_name": "background-remove",
  "inputs": {
    "image": "https://example.com/product-photo.jpg"
  }
}

Image to Video (`image-to-video`)

Turn a single image into a short motion clip with fashion-tailored camera work and model movements.

Parameter	Type	Required	Description
`image`	string	Yes	Source image to animate
`prompt`	string	No	Motion guidance (keep short, e.g. "raising hand to touch face")
`negative_prompt`	string	No	Cues to avoid undesirable motion
`duration`	number	No	`5` or `10` seconds (default: `5`)
`resolution`	string	No	`480p`, `720p`, `1080p` (default: `1080p`)

{
  "model_name": "image-to-video",
  "inputs": {
    "image": "https://example.com/model-photo.jpg",
    "duration": 5,
    "resolution": "1080p"
  }
}

Long Processing Time

Image-to-video takes 1-3 minutes. Use the /run + /status pattern for this model, as /subscribe has a 2-minute timeout.

Image Inputs

All image parameters accept either:

URL: Publicly accessible image URL with correct Content-Type headers
Base64: Data URI with format prefix, e.g. data:image/jpg;base64,/9j/4AAQ...

Base64 Format

Base64 images must include the proper prefix: data:image/jpg;base64,<YOUR_BASE64> or data:image/png;base64,<YOUR_BASE64>. Omitting the prefix will cause an ImageLoadError.

Credits

Model	Base Cost	With `face_reference`	Multiple Outputs
`tryon-v1.6`	3 credits	—	× num_samples
`product-to-model`	3 credits	5 credits	× num_images
`face-to-model`	3 credits	—	—
`model-create`	3 credits	5 credits	× num_images
`model-variation`	3 credits	—	—
`model-swap`	3 credits	5 credits	× num_images
`reframe`	2 credits	—	× num_images
`background-change`	2 credits	—	—
`background-remove`	2 credits	—	—
`image-to-video`	5 credits	—	—

Failed predictions do not consume credits.

Rate Limits

Endpoint	Limit
`/api/v1/fashion/run`	30 requests per 60 seconds
`/api/v1/fashion/subscribe`	30 requests per 60 seconds
`/api/v1/fashion/status/:id`	40 requests per 10 seconds

Rate limit headers are included in every response:

X-RateLimit-Limit — Max requests allowed
X-RateLimit-Remaining — Requests remaining in window
X-RateLimit-Reset — Unix timestamp when window resets

Error Handling

API-Level Errors

Returned before a prediction is created (no prediction ID):

Code	Error	Solution
400	Validation failed	Check `model_name` and required `inputs` fields
401	Invalid API key	Verify your `Authorization` or `X-API-Key` header
429	Rate limit exceeded	Throttle requests, check `Retry-After` header
503	Service unavailable	Fashion generation service is not configured on the server

Runtime Errors

Returned when polling /status/:id with status: "failed":

Error Name	Cause	Solution
`ImageLoadError`	Cannot fetch/decode input image	Use public URLs with image Content-Type, or valid base64 with prefix
`ContentModerationError`	Input violated content policies	Adjust input or set `moderation_level` to `permissive`
`PoseError`	Cannot detect body pose	Improve image quality, ensure full body is visible
`InputValidationError`	Invalid parameter values	Check parameter types and enum values
`PipelineError`	Internal processing failure	Retry the request
`ThirdPartyError`	Upstream service failure	Retry with backoff

Failed = Free

Failed predictions do not consume credits. It's safe to retry.

Fashion API

On this page