/demo-automatic1111-sd-webui

Primary LanguagePythonMIT LicenseMIT

Banana Implementation

This repository is the implementation of stable-diffusion-webui

Endpoints

At the moment only one endpoint is implemented.

  • /txt2img is the implementation of stable-diffusion-webui /sdapi/v1/txt2img api.

    Json Input

    {
    "body": {
        "params": {
            "prompt": "CAT",
            "negative_prompt": "low quality",
            "steps": 25,
            "sampler_name": "Euler a",
            "cfg_scale": 7.5,
            "seed": 42,
            "batch_size": 1,
            "n_iter": 1,
            "width": 400,
            "height": 400,
            "tiling": 'false',
        }
    }
}

    OUTPUT:

    {
    "output": "<base64 image>"
}

  • img2img is the implementation of /sdapi/v1/img2img api

    JSON INPUT

    img2img_inputs = {
    "params": {
        "prompt": "blue banana",
        "negative_prompt": "cartoonish, low quality",
        "steps": 25,
        "sampler_name": "Euler a",
        "cfg_scale": 7.5,
        "denoising_strength": 0.7,
        "seed": 42,
        "batch_size": 1,
        "n_iter": 1,
        "width": 768,
        "height": 768,
        "tiling": "false",
        "init_images": [
            "b64images"
        ]
    }
}

Client Implementation in Python

import banana_dev as client

my_model = client.Client(
    api_key="[API_KEY]",
    model_key="[MODEL_KEY]",
    url="[MODEL_URL]",
)

txt2img_inputs = {
    "params": {
        "prompt": "Nurse in scrub dress with her hair tied",
        "negative_prompt": "low quality",
        "steps": 25,
        "sampler_name": "Euler a",
        "cfg_scale": 7.5,
        "seed": 42,
        "batch_size": 1,
        "n_iter": 1,
        "width": 400,
        "height": 400,
        "tiling": "false",
    }
}

img2img_inputs = {
    "params": {
        "prompt": "blue banana",
        "negative_prompt": "cartoonish, low quality",
        "steps": 25,
        "sampler_name": "Euler a",
        "cfg_scale": 7.5,
        "denoising_strength": 0.7,
        "seed": 42,
        "batch_size": 1,
        "n_iter": 1,
        "width": 768,
        "height": 768,
        "tiling": "false",
        "init_images": [
            "b64images"
        ]
    }
}



txt2img_result, meta = my_model.call("/txt2img", txt2img_inputs)
img2img_result, meta = my_model.call("/img2img", img2img_inputs)

print(txt2img_inputs, img2img_result)