POST
/
train-model
cURL
curl --request POST \
  --url https://tryiton-webapp-git-dev-try-it-on.vercel.app/api/key/train-model \
  --header 'Content-Type: application/json' \
  --data '{
  "modelId": "<string>",
  "gender": "man",
  "modelApi": "modal",
  "styles": "<string>",
  "customizations": "<string>",
  "imageUrls": [
    "<string>"
  ]
}'
{}

​
The ‘styles’ parameter

This parameter corresponds to the “Styles” step when customizing a freshly bought model in the studio. For the styles parameter pass down at least one of the styles from the ones available the chosen gender. You can use the previous endpoint to get all the available styles per gender. If you’d like to upload your own style, do note that you have to first get the key of the uploaded style using the image upload endpoint, and then ensure you use it within 24 hours, as it will be deleted automatically thereafter.

​
The ‘customizations’ parameter

This parameter corresponds to the “Customizations” step when customizing a freshly bought model in the studio. It should contain defining features of the user, such as eye color, hair color, haircut style, glasses or beard. Glasses or beard should only be mentioned if the subject’s photos contain those features. No need to specify “no beard” or “no glasses”, however. These features should be separated by comma and space (, ). Lastly, don’t overdo it because it will limit the AI’s creativity in generating photos, or potentially confuse concepts. Here is a good example for a male subject:
wearing glasses, thick beard, mohawk haircut

​
The ‘imageUrls’ parameter

You must upload square images with a resolution of 640x640 (you can pad them with a white background if the subject crops are not square). If the images are not square, the AI will crop them to a square, potentially cutting off important parts of the subject. The images should be real photos containing the same subject and no other people in them. You can check the guidelines presented when customizing a freshly bought model for more details. Regarding the location of these images, as long as CORS is enabled on the domain hosting the images, we’ll be able to access them. To be extra safe, you can also use our image upload endpoint to store your images and use image urls that will certainly work.

Body

application/json

Response

200
application/json

Successful response

The response is of type object.