Generate images

The image generation request, every field it takes, the sizes you can ask for, and what comes back.

POST /image/generate creates an image task. Send a model id and a prompt. Everything else is optional.

import Mynth from "@mynthio/sdk";

const mynth = new Mynth();

const task = await mynth.image.generate({
  model: "black-forest-labs/flux.2-pro",
  prompt: "A lighthouse at dusk, film grain",
  size: "landscape",
});

console.log(task.urls[0]);

The SDK and the CLI wait and return the finished task. The REST call returns at once with a task id:

{
  "data": {
    "taskId": "tsk_01KE7XWWEQ4MCGWKBQKJ1G47RP",
    "estimatedCost": "0.03",
    "access": { "publicAccessToken": "pat_eyJhbGciOiJIUzI1NiJ9..." }
  }
}

Collect the result by polling or with a webhook.

Fields#

Every image model takes this body.

Field	Default	Notes
`model`	`auto`	A catalog id. Always pass one. `auto` is experimental
`prompt`	required	1 to 8192 characters
`negative_prompt`	none	Up to 8192 characters. Dropped without an error if the model has none
`count`	`1`	1 to 20. Each image succeeds or fails on its own
`size`	`auto`	See sizes
`inputs`	none	Up to 20 input images. See image to image
`magic_prompt`	`false`	Rewrite the prompt first. See enhance prompts
`rating`	none	`true`, or a rating object. See rate images
`output.format`	the provider's	`png`, `jpg`, or `webp`
`destination`	none	A destination name. See destinations
`webhook`	registered ones	Per-request URLs. See webhooks
`metadata`	none	A JSON object up to 2048 bytes, returned on the task and in webhooks
`access`	token issued	`{ "pat": { "enabled": false } }` skips the browser token

Fields the schema does not know are dropped without an error. If a setting seems to have no effect, check the task's request, which shows the body as it was accepted.

Sizes#

You ask for a shape, not pixels, and Mynth picks the closest size the model serves. Omit size, or send "auto", and Mynth picks a ratio that suits the prompt.

`size`	Result
`square`, `portrait`, `landscape`, `portrait_tall`, `landscape_wide`	1:1, 2:3, 3:2, 9:16, 16:9
`1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`, `2:1`, `1:2`	that ratio
Any ratio with `_4k`, such as `16:9_4k`	that ratio at 4k
`{ "type": "aspect_ratio", "aspectRatio": "16:9", "scale": "4k" }`	the same, as an object

A ratio the model has no preset for snaps to the closest one it serves. A 4k size costs perImage["4k"] and needs a model that has it. On a model without 4k the task fails with CAPABILITY_NOT_SUPPORTED.

What comes back#

A completed task has one entry in result.images per requested image:

{
  "model": "black-forest-labs/flux.2-pro",
  "images": [
    {
      "status": "success",
      "id": "img_V1StGXR8Z5jdHi6BmyT0sC1pQ2rN4wLk",
      "url": "https://cdn.mynth.io/images/img_V1StGXR8Z5jdHi6BmyT0sC1pQ2rN4wLk.webp",
      "mynth_url": "https://cdn.mynth.io/images/img_V1StGXR8Z5jdHi6BmyT0sC1pQ2rN4wLk.webp",
      "size": "1536x1024",
      "format": "webp"
    }
  ]
}

Check status on every image. One image failing does not fail the task. See tasks.
url is where the file ended up: your storage when you named a destination, otherwise the same as mynth_url.
size is the delivered file in {width}x{height} pixels.
result.model is the model that ran.
The file is served for 7 days. See file lifetimes.

Every field of the result is on the task object.

Price#

The create response holds estimatedCost, and you pay only for the images that succeed. Pricing has the formula, and estimate cost prices a body without generating.

Next steps#

Choosing an image model: what to check before you pin an id.
Image to image: send input images.
Poll for results: collect the task over REST.