1. Input batches
To create a finetune on custom uploaded data, the first thing you need to do is upload your input images. A group of input images is called an input batch, and is essentially a folder.
POST /input-batches
Request
Your request must be encoded as multipart/form-data.
{
"images": [File1, File2, ...File20],
"label": "my selfies"
}images (required)
Your input images. Please resize them to 512px x 512px and save them as JPGs before uploading. Minimum: 1 image. Maximum: 30 images.
We recommend uploading a diverse range of face shots and upper-body shots, taken in different locations, different costumes, different lighting conditions, and with different cameras. If the only thing consistent between your images is the subject, then the model will learn better and produce higher quality generations.
label
Optionally add a label to your input batch, for easier management of your image data.
We recommend you resize your images to 512px x 512px before uploading. If you upload images with random sizes, then a plain crop will be used to resize them, and the results may be disappointing.
Response
{
"id": "65c82e92-eed9-462b-829d-5807aadda53f",
"label": "my selfies",
"created_at": "2022-11-02T22:33:54.348215+00:00",
"num_successful_uploads": 20,
"num_failed_uploads": 0
}GET /input-batches/{input_batch_id}
Response
{
"id": "65c82e92-eed9-462b-829d-5807aadda53f",
"label": "my selfies",
"created_at": "2022-11-02T22:33:54.348215+00:00"
}