Images

Generate, edit, and create variations of images via DALL-E, Stable Diffusion, Flux, Imagen, and Ideogram models.

Overview

The Images API exposes three operations behind a single OpenAI-compatible interface:

Generation (/v1/images/generations) -- create an image from a text prompt
Edit (/v1/images/edits) -- inpaint/extend an existing image
Variations (/v1/images/variations) -- generate variations of a source image

You can route to DALL-E 3, Stable Diffusion 3, Flux, Imagen 3, Ideogram, or Recraft by changing the model identifier.

Generation

POST https://api.allroutes.ai/v1/images/generations

Request Body

Field	Type	Required	Description
`model`	string	Yes	Image model (e.g., `dall-e-3`, `flux-pro-1.1`, `stable-diffusion-3.5-large`)
`prompt`	string	Yes	Text description (max 4000 characters for DALL-E 3)
`size`	string	No	`1024x1024` (default), `1024x1792`, `1792x1024`, `512x512`
`n`	integer	No	Number of images (1 for DALL-E 3, up to 10 for SD/Flux)
`quality`	string	No	`standard` (default) or `hd`
`style`	string	No	`vivid` (default) or `natural` (DALL-E 3 only)
`response_format`	string	No	`url` (default) or `b64_json`
`negative_prompt`	string	No	Things to avoid (SD/Flux only)
`seed`	integer	No	Reproducible generation (SD/Flux only)
`aspect_ratio`	string	No	Flux/SD-style aspect ratio: `1:1`, `16:9`, `4:3`

Example

curl https://api.allroutes.ai/v1/images/generations \
  -H "Authorization: Bearer allroutes_sk_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "dall-e-3",
    "prompt": "A cyberpunk city at sunset, photorealistic, ultra detailed",
    "size": "1024x1792",
    "quality": "hd",
    "n": 1
  }'

Response

{
  "created": 1714000000,
  "data": [
    {
      "url": "https://cdn.allroutes.ai/img/abc123.png",
      "revised_prompt": "A cyberpunk metropolis at golden hour, ..."
    }
  ]
}

URLs returned in url mode expire after 1 hour. Use response_format: "b64_json" for permanent storage in your own infrastructure.

Edit (Inpainting / Outpainting)

POST https://api.allroutes.ai/v1/images/edits

Multipart upload with an image and a mask. Transparent pixels in the mask indicate regions to regenerate.

curl https://api.allroutes.ai/v1/images/edits \
  -H "Authorization: Bearer allroutes_sk_..." \
  -F image=@photo.png \
  -F mask=@mask.png \
  -F prompt="A cat sitting where the dog used to be" \
  -F model="dall-e-2" \
  -F size="1024x1024"

Supported models: dall-e-2, flux-pro-edit, stable-diffusion-inpaint.

Variations

POST https://api.allroutes.ai/v1/images/variations

Generate variations of a source image while preserving subject, composition, and style.

curl https://api.allroutes.ai/v1/images/variations \
  -H "Authorization: Bearer allroutes_sk_..." \
  -F image=@source.png \
  -F model="dall-e-2" \
  -F n=4 \
  -F size="1024x1024"

Currently supported: dall-e-2, stable-diffusion-3-variation.

SDK Examples

Python

# Generation
result = client.images.generate(
    model="dall-e-3",
    prompt="A cyberpunk city at sunset, photorealistic",
    size="1024x1792",
    quality="hd",
)
url = result.data[0].url
print(url)

# Edit
with open("photo.png", "rb") as img, open("mask.png", "rb") as mask:
    edit = client.images.edit(
        model="dall-e-2",
        image=img,
        mask=mask,
        prompt="Replace the dog with a cat",
        size="1024x1024",
    )

Node.js

const result = await client.images.generate({
  model: "dall-e-3",
  prompt: "A cyberpunk city at sunset, photorealistic",
  size: "1024x1792",
  quality: "hd",
});
console.log(result.data[0].url);

Supported Models

Model	Provider	Strength
`dall-e-3`	OpenAI	Best prompt adherence, single image only
`dall-e-2`	OpenAI	Cheap, supports edits and variations
`flux-pro-1.1`	Black Forest Labs	Highest fidelity, photorealistic
`flux-schnell`	Black Forest Labs	Fastest, good for ideation
`stable-diffusion-3.5-large`	Stability AI	Open weights, customizable
`imagen-3`	Google	Strong typography and text rendering
`ideogram-v2`	Ideogram	Best for posters/logos with text
`recraft-v3`	Recraft	Vector and design-style outputs

Pricing

Image pricing is per-image, not per-token:

Model	Cost per image (1024x1024 standard)
`dall-e-3`	$0.040
`dall-e-3` (HD)	$0.080
`dall-e-2`	$0.020
`flux-pro-1.1`	$0.040
`flux-schnell`	$0.003
`stable-diffusion-3.5-large`	$0.065
`imagen-3`	$0.040

See Models for live pricing.

Best Practices

Use revised_prompt -- DALL-E 3 rewrites prompts for safety; the field reveals what was actually used so you can iterate
Set seed for SD/Flux -- reproducibility helps with prompt iteration and A/B comparisons
Pick by strength -- DALL-E 3 for prompt adherence; Flux Pro for photorealism; Ideogram for text-in-image
Cache by prompt -- the gateway's exact-match cache covers identical (prompt, model, size) requests, saving the per-image cost on retries
Download promptly -- url responses expire after 1 hour; download or use b64_json for archival

Troubleshooting

Symptom	Cause	Fix
`400: prompt too long`	Over 4000 chars	Trim or summarize the prompt
`400: content_policy_violation`	Safety filter triggered	Reword or switch to a less-restricted model
Faces look unrealistic	Generic SD weights	Use `flux-pro-1.1` or DALL-E 3 HD
Output text is garbled	Most models struggle with text	Use `imagen-3` or `ideogram-v2`