API img2img

[sayyoume]

http://127.0.0.1:7860/sdapi/v1/img2img api
I would like to know which parameters are used to identify inpaint /sketch /inpaint inpaint / inpain upload

[huotarih]

me too

[OlegXio]

Окей, долго работал над этим, ловите:

import os
import requests
import io
import base64
import random
import json
from PIL import Image, PngImagePlugin
api = 'http://127.0.0.1:7860'
try:
    os.makedirs('diffimages')
except:
    pass
def img2img(api, text, steps, image_path):
    api_url = f"{api}/sdapi/v1/img2img"
    with open(image_path, 'rb') as file:
        image_data = file.read()
    encoded_image = base64.b64encode(image_data).decode('utf-8')
    payload = {
        "init_images": [encoded_image],
        'prompt' : text,
        "steps": steps,
    }
    response = requests.post(api_url, json=payload)
    name = 'GENimg2img_'
    for i in range(random.randint(15, 25)):
        name += random.choice('QAZXfrSWEDCVFRTqazxswgbnhyujmkiolpGBNHYUJedcvtMKIOLP')
    print(name)
    if response.status_code == 200:
        response_data = response.json()
        encoded_result = response_data["images"][0]
        result_data = base64.b64decode(encoded_result)
        output_path = f"diffimages/{name}.jpg"
        with open(output_path, 'wb') as file:
            file.write(result_data)
        return name
    else:
        print("Ошибка при выполнении запроса:", response.text)
        return None

[techykajal]

Hey @OlegXio , Can you please explain me what kind of input images you have used? I'm trying to automate the process of image to image generation by inpainting specific region.

• One thing that I could think of sending whole image + sending mask of the region where I would like to inpaint and further generate that area using prompt.
  I'm skeptical that .csv file may be required for this task but not sure how to include the images in it.

Please help.
Thanks.

[OlegXio]

I just used jpg/PNG images and a text query, and in the code I provided there is no function for selecting the image area (inpaint), and in my opinion there is no api either.

[levelingup]

hi @sayyoume did you ever find out?

[vual]

me too

[OlegXio]

see above, I gave the working code

[sunatte-saad]

you will have to create a mask and its image and then send it to img2img

[ym-ctw]

you will have to create a mask and its image and then send it to img2img
How to confirm that you are using inpaint sketch not inpaint in this way?

[sunatte-saad]

It is only different in UI, in the code, it is same

[w-e-w]

img2img Sketch Inpaint Inpaint sketch Inpaint upload
all these mode doesn't actually exists
there's no difference in terms of what stable division do, the difference is what input and mask you gave it

the API is technically just Inpaint upload
img2img is Inpaint upload but mask is blank
Inpaint is Inpaint upload with a bask that is not blank
Sketch is img2img except your input image have extra strokes added to it
Inpaint sketch is Sketch but with a mask that corresponds to the strokes

this is the code in webUI that does the processing when you're using it via the webUI

stable-diffusion-webui/modules/img2img.py

Lines 154 to 176 in cb5b335

	if mode == 0: # img2img
	image = init_img
	mask = None
	elif mode == 1: # img2img sketch
	image = sketch
	mask = None
	elif mode == 2: # inpaint
	image, mask = init_img_with_mask["image"], init_img_with_mask["mask"]
	mask = processing.create_binary_mask(mask)
	elif mode == 3: # inpaint sketch
	image = inpaint_color_sketch
	orig = inpaint_color_sketch_orig or inpaint_color_sketch
	pred = np.any(np.array(image) != np.array(orig), axis=-1)
	mask = Image.fromarray(pred.astype(np.uint8) * 255, "L")
	mask = ImageEnhance.Brightness(mask).enhance(1 - mask_alpha / 100)
	blur = ImageFilter.GaussianBlur(mask_blur)
	image = Image.composite(image.filter(blur), orig, mask.filter(blur))
	elif mode == 4: # inpaint upload mask
	image = init_img_inpaint
	mask = init_mask_inpaint
	else:
	image = None
	mask = None

with API you have to do it yourself

and batch mode is just multiple jobs