curl --request POST \
--url https://api.minimax.io/v1/h3_image_context_ir \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"model": "MiniMax-H3-Image",
"prompt": "A ginger cat sunbathing on a windowsill, warm afternoon light, film look"
}
'{
"created": 1794023743,
"revised_prompt": "A ginger tabby cat lounges lazily on a wooden windowsill, basking in the sun. Warm golden afternoon light slants in through the window, casting soft highlights on its fur and the sill. The curtains and room behind are slightly out of focus, and the whole image has a warm film look with fine grain and warm tones.",
"aspect_ratio": "3:2",
"usage": {
"total_tokens": 3489,
"input_tokens": 3489
},
"request_id": "021792308678142a2c883cf675b9804d"
}H3-Image-Context-IR
Run MiniMax-H3-Image prompt enhancement on its own and get the enhanced prompt synchronously, without generating images.
curl --request POST \
--url https://api.minimax.io/v1/h3_image_context_ir \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"model": "MiniMax-H3-Image",
"prompt": "A ginger cat sunbathing on a windowsill, warm afternoon light, film look"
}
'{
"created": 1794023743,
"revised_prompt": "A ginger tabby cat lounges lazily on a wooden windowsill, basking in the sun. Warm golden afternoon light slants in through the window, casting soft highlights on its fur and the sill. The curtains and room behind are slightly out of focus, and the whole image has a warm film look with fine grain and warm tones.",
"aspect_ratio": "3:2",
"usage": {
"total_tokens": 3489,
"input_tokens": 3489
},
"request_id": "021792308678142a2c883cf675b9804d"
}images, the prompt is enhanced as text-to-image; with reference images, it is enhanced as an image edit. Use it to preview and revise the enhanced prompt, or to call enhancement as a separate step in your own workflow.
Supported model: MiniMax-H3-Image; MiniMax-H3-Image-Highspeed is not supported by this endpoint and returns 400. The server waits up to 600 seconds; set the client timeout to at least 600 seconds.
usage.input_tokens; no output tokens are billed separately. Failed requests (including those that fail the content safety check) are not billed.revised_prompt as the prompt of a Text-to-Image request with disable_prompt_enhanced: true and the aspect_ratio from the response (or your own size / aspect_ratio); the generation request is then not billed for prompt enhancement again. Image edits always apply prompt enhancement, so passing the result to Image Edit enhances and bills it again.Authorizations
HTTP: Bearer Auth
- Security Scheme Type: http
- HTTP Authorization Scheme:
Bearer API_key, used to verify account information, can be found in Account Management>API Keys.
Body
Model name. Supported value: MiniMax-H3-Image
MiniMax-H3-Image Text description to enhance, up to 32000 characters (counted as Unicode characters, excluding leading and trailing whitespace); an empty or whitespace-only prompt returns 400.
32000Reference images, 1 to 16. When provided, the prompt is enhanced as an image edit using the reference images; when omitted, it is enhanced as text-to-image.
Reference image limits (total request body ≤ 64 MiB; Base64-encoded data is about 4/3 of the original file size; if the request body would exceed the limit, use public URLs or mm_file://)
| Item | Limit |
|---|---|
| Format | JPG, JPEG, PNG, WEBP, HEIC, HEIF |
| Single file size | ≤ 64 MiB |
| Width/height range | [128, 8192] px |
| Aspect ratio (long/short side) | ≤ 4:1 |
| Count | 1 to 16 |
1 - 16 elementsShow child attributes
Show child attributes
Target output size, default auto. Same forms as the generation endpoints; use one of two forms; see Output Size.
WIDTHxHEIGHT: e.g.2720x1536; image-edit enhancement describes the scene in this width-to-height ratio. Area and aspect ratio limits are the same as the generation endpoints; otherwise 400 is returned. In this form, do not setaspect_ratioorarea(not even toauto); otherwise 400 is returnedautoor omitted: the aspect ratio is determined byaspect_ratio
Size parameters only affect image-edit enhancement; text-to-image enhancement does not read them.
Target aspect ratio, default auto. Can only be set when size is omitted or auto; setting it while size is WIDTHxHEIGHT returns 400.
auto: decided by prompt enhancement; seeaspect_ratioin the responseX:Y: X and Y are integers from 1 to 300 and the ratio is between 1:3 and 3:1 (bounds inclusive), e.g.16:9,3:4. Image-edit enhancement describes the scene in this ratio, and the response does not includeaspect_ratio
"16:9"
Area level: 1K, 2K or 4K, default 4K. Can only be set when size is omitted or auto; setting it while size is WIDTHxHEIGHT returns 400. It does not affect the enhancement result and lets you reuse the same parameters as the generation request.
1K, 2K, 4K Response
Enhancement succeeded; returns the enhanced prompt.
Time the response was generated, as a Unix timestamp in seconds.
The enhanced prompt.
Aspect ratio chosen by prompt enhancement, in X:Y format and between 1:3 and 3:1. Returned only when the ratio is decided by enhancement (size omitted or auto, and aspect_ratio omitted or auto).
"3:2"
Token usage of this request, used for billing. The cost of prompt enhancement is converted into input text tokens; no output tokens are billed separately.
Show child attributes
Show child attributes
Request ID (the trace ID). Provide it when reporting an issue.