Image Generation

Generate images from text prompts or edit existing images.

Available Endpoints

Provider	Endpoint	Type	Model
Stable Diffusion	/v2/automatic1111/sdapi/v1/txt2img	T2I	Various SD models
Flux	/v2/image/flux/generate	T2I	black-forest-labs/FLUX-2-pro
Gemini	/v2/image/gemini-3/generate	T2I	google/gemini-3-pro-T2I
GPT Image	/v2/image/gpt-image/generate	T2I	openai/gpt-image-1.5-T2I
GPT Image	/v2/image/gpt-image/edit	I2I	openai/gpt-image-1.5-I2I
Qwen	/v2/image/qwen/generate	T2I	Qwen/qwen-image
Qwen	/v2/image/qwen/edit	I2I	Qwen/qwen-image-edit

---

Stable Diffusion

POST /v2/automatic1111/sdapi/v1/txt2img

Parameter	Type	Default	Description
prompt	string	required	Image description
negative_prompt	string	-	What to avoid
model	string	Realistic_Vision_V5.1.safetensors	SD model
steps	int	20	Denoising steps
width	int	512	Image width
height	int	512	Image height
cfg_scale	float	7	Prompt adherence

Models: Realistic_Vision_V5.1.safetensors, illustriousRealismBy_v10VAE.safetensors, stable-diffusion-v1-5-pruned.safetensors

Response: images (base64 array)

curl -X POST https://relay.opengpu.network/v2/automatic1111/sdapi/v1/txt2img \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "mountain landscape at sunset", "steps": 30}'

---

Flux

POST /v2/image/flux/generate

Parameter	Type	Default	Description
prompt	string	required	Image description
width	int	1024	Width (min 64, multiple of 32)
height	int	1024	Height (min 64, multiple of 32)
seed	int	-	Random seed
safety_tolerance	int	2	Safety level (0=strictest, 6=least)
output_format	string	png	png or jpeg
input_image	string	-	Reference image 1 (base64)
input_image_2	string	-	Reference image 2 (base64)
input_image_3	string	-	Reference image 3 (base64)
input_image_4	string	-	Reference image 4 (base64)

Response: images (base64 array)

curl -X POST https://relay.opengpu.network/v2/image/flux/generate \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "A red fox in autumn forest", "width": 1024, "height": 1024}'

---

Gemini

POST /v2/image/gemini-3/generate

Parameter	Type	Default	Description
prompt	string	required	Image description
aspect_ratio	string	1:1	1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9
image_size	string	1K	1K, 2K, 4K

Response: images (base64 array)

curl -X POST https://relay.opengpu.network/v2/image/gemini-3/generate \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "A serene mountain landscape", "aspect_ratio": "16:9", "image_size": "2K"}'

---

GPT Image

Generate (T2I)

POST /v2/image/gpt-image/generate

Parameter	Type	Default	Description
prompt	string	required	Image description
size	string	1024x1024	1024x1024, 1536x1024, 1024x1536, auto
quality	string	auto	auto, low, medium, high
n	int	1	Number of images (1-4)

Response: images (URL array)

curl -X POST https://relay.opengpu.network/v2/image/gpt-image/generate \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "A sunset over mountains", "size": "1536x1024", "quality": "high"}'

Edit (I2I)

POST /v2/image/gpt-image/edit

Parameter	Type	Default	Description
prompt	string	required	Edit instructions
image	string	required	Source image (base64)
mask	string	-	Edit mask (base64, white=edit area)
size	string	1024x1024	Output size
quality	string	auto	auto, low, medium, high
n	int	1	Number of images (1-4)

Response: images (URL array)

curl -X POST https://relay.opengpu.network/v2/image/gpt-image/edit \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "Add a rainbow to the sky", "image": "<base64>"}'

---

Qwen

Generate (T2I)

POST /v2/image/qwen/generate

Parameter	Type	Default	Description
prompt	string	required	Image description
size	string	1024x1024	Output size (WxH format)
seed	int	-	Random seed

Response: urls (URL array, expire after 7 days)

curl -X POST https://relay.opengpu.network/v2/image/qwen/generate \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "A beautiful sunset over mountains"}'

Edit (I2I)

POST /v2/image/qwen/edit

Parameter	Type	Default	Description
prompt	string	required	Edit instructions
image	string	required	Source image (URL or base64)
size	string	1024x1024	Output size (WxH format)
seed	int	-	Random seed

Response: urls (URL array, expire after 7 days)

curl -X POST https://relay.opengpu.network/v2/image/qwen/edit \
  -H "X-API-Key: YOUR_KEY" \
  -d '{"prompt": "Convert to pencil sketch", "image": "https://example.com/photo.jpg"}'

---

Async Mode

All image endpoints support "async": true for longer generations.

{"prompt": "...", "async": true}

Response (202):

{
  "task_address": "direct:abc12345-...",
  "status": "running",
  "poll_url": "/v2/tasks/direct:abc12345-..."
}

Poll: GET /v2/tasks/{task_address}

---

Notes

• Timeout: 5 minutes (use async for complex prompts)

• Output: Some providers return base64, others return URLs

• Credits charged on completion