curl --request POST \
--url https://api.tikway.ai/v1/images/generations \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"model": "diffusion-4-5-full",
"prompt": "A serene Japanese garden at dawn, mist above a koi pond, cinematic lighting, highly detailed",
"negative_prompt": "lowres, blurry, text, watermark",
"size": "1024x1536",
"n": 1,
"steps": 28,
"scale": 5,
"sampler": "Euler Ancestral",
"noise_schedule": "karras"
}
'{
"created": 1713833628,
"background": "opaque",
"data": [
{
"b64_json": "...",
"index": 0
}
],
"output_format": "png",
"size": "1024x768"
}Image generation
Use POST /v1/images/generations for text-to-image and image-to-image with diffusion-4-5-full or diffusion-4-5-curated. Supply source images and style or character references through images. n supports 1–4 outputs; only one ordinary source image is allowed. vibe_transfer and character_reference cannot be combined.
curl --request POST \
--url https://api.tikway.ai/v1/images/generations \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"model": "diffusion-4-5-full",
"prompt": "A serene Japanese garden at dawn, mist above a koi pond, cinematic lighting, highly detailed",
"negative_prompt": "lowres, blurry, text, watermark",
"size": "1024x1536",
"n": 1,
"steps": 28,
"scale": 5,
"sampler": "Euler Ancestral",
"noise_schedule": "karras"
}
'{
"created": 1713833628,
"background": "opaque",
"data": [
{
"b64_json": "...",
"index": 0
}
],
"output_format": "png",
"size": "1024x768"
}Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Body
NovelAI image generation Request Body. The unified reference image uses images; the gateway is responsible for converting to model parameters.
Required. Use a model ID or available alias returned by Get Models.
1"diffusion-4-5-full"
"diffusion-4-5-curated"
Required. Describe the image to generate.
1 - 32000Optional reference images for image-to-image generation. Omit for text-to-image generation. Image limits depend on the selected model.
16Hide child attributes
Hide child attributes
1image or null is the image source image; the remaining values represent the NovelAI reference image type.
image, vibe_transfer, character_reference, null Reference strength. The maximum value of the plot is 0.99.
0 <= x <= 1Fidelity. The graph is converted to noise = 1 - fidelity.
0 <= x <= 1Only used by character_reference.
character, style, character&style Negative cue words.
Number of generated images. Allowed counts depend on the selected model.
1 <= x <= 4Output size as WIDTHxHEIGHT, or auto for the model default. Supported sizes depend on the selected model.
^[1-9][0-9]*x[1-9][0-9]*$Random seed.
Number of sampling steps.
x >= 1Cue words guide intensity.
Quality switch.
Euler Ancestral, Euler, DPM++ 2S Ancestral, DPM++ 2M SDE, DPM++ 2M, DPM++ SDE karras, exponential, polyexponential Response format: url returns image URLs; b64_json returns Base64 image data.
b64_json, url Response
The Unix timestamp (in seconds) when the gateway completed image conversion.
Background settings inferred from unified inbound requests; NovelAI raw responses are typically not echoed.
transparent, opaque The format inferred from the downloaded image MIME type or request output_format.
png, jpeg, webp Unified quality field; if quality is requested, the gateway will map and echo it.
low, medium, high The size of the original request; NovelAI's original response is usually not echoed.