SEND JOB TO COMFY UI

[enzyme69]

Install ComfyUI using PINOKIO AI Browser. Having it running as server backend for AI Generative.

I am using Mac Mini M2 Pro 32 GB. Seems good base model, not the fastest, but silent and speedy anough. Make sure you have 2 TB or more.

# SDXL Turbo model is used, but LoRA LCM was also fast

import bpy
import json
import requests
import random

# ComfyUI server address
comfyui_address = "http://127.0.0.1:8188"

def queue_prompt(prompt_workflow):
    p = {"prompt": prompt_workflow}
    
    # Convert prompt workflow to JSON
    workflow_json = json.dumps(p).encode('utf-8')

    # Send POST request to ComfyUI server
    response = requests.post(comfyui_address + "/prompt", data=workflow_json)

    # Check if request was successful
    if response.status_code != 200:
        print("Error sending prompt to ComfyUI:", response.status_code)
        print(response.text)
    else:
        print("Prompt successfully sent to ComfyUI")

# Read workflow API data from file and convert it into dictionary
prompt_workflow = json.load(open("/Users/jimmygunawan/Downloads/sdxl_turbo_api.json"))

# Create a list of prompts

prompt_list = []

# Access the text block and append for ComfyUI
# Get the text datablock
text_datablock = bpy.data.texts['MonkeyRobot']

# Get the text content
text_content = text_datablock.lines

if text_datablock:
    # Read the contents of the text block
    for i in text_content:
        print(i.body)
        prompt_list.append(i.body)
else:
    print(f"Text block '{text_block_name}' not found.")

#prompt_list = []
#prompt_list.append("photo of a bearded man Mario wearing a red hat sitting alone in a cafe")
#prompt_list.append("oil painting of a llama standing in the middle of a busy street")
#prompt_list.append("water colour art of a fish sitting in a tree")
#prompt_list.append("beautiful mushroom glass bottle landscape, purple galaxy bottle")
#prompt_list.append("princess peach on abandoned castle reading book at night, fireflies")
#prompt_list.append("head totem of monkey head, made of mecha garbage recycle, in the middle of empty park")

# Give some easy-to-remember names to the nodes
prompt_pos_node = prompt_workflow["2"]
seed_node = prompt_workflow["4"]
#save_image_node = prompt_workflow["1"]
#empty_latent_img_node = prompt_workflow["1"]

# Set image dimensions and batch size in EmptyLatentImage node
#empty_latent_img_node["inputs"]["width"] = 512
#empty_latent_img_node["inputs"]["height"] = 640
#empty_latent_img_node["inputs"]["num_images"] = 1

# For every prompt in prompt_list...
for index, prompt in enumerate(prompt_list):
    # Set the text prompt for positive CLIPTextEncode node
    prompt_pos_node["inputs"]["text"] = prompt + " made of garbage, waste recycles, wabisabi, imperfect"

    # Set a random seed in KSampler node
    seed_node["inputs"]["noise_seed"] = random.randint(1, 18446744073709551614)

    # Set filename prefix to be the same as prompt
    # (truncate to first 100 chars if necessary)
    #fileprefix = prompt
    #if len(fileprefix) > 100:
    #    fileprefix = fileprefix[:100]

    #save_image_node["inputs"]["filename_prefix"] = fileprefix

    # Everything set, add entire workflow to queue.
    queue_prompt(prompt_workflow)

[enzyme69]

sdxl_turbo_api.json.zip

{
"1": {
"inputs": {
"ckpt_name": "sd_xl_turbo_1.0_fp16.safetensors"
},
"class_type": "CheckpointLoaderSimple"
},
"2": {
"inputs": {
"text": "oil painting style of ",
"clip": [
"1",
1
]
},
"class_type": "CLIPTextEncode"
},
"3": {
"inputs": {
"text": "",
"clip": [
"1",
1
]
},
"class_type": "CLIPTextEncode"
},
"4": {
"inputs": {
"add_noise": true,
"noise_seed": 340269611110808,
"cfg": 1,
"model": [
"1",
0
],
"positive": [
"2",
0
],
"negative": [
"3",
0
],
"sampler": [
"13",
0
],
"sigmas": [
"9",
0
],
"latent_image": [
"10",
0
]
},
"class_type": "SamplerCustom"
},
"9": {
"inputs": {
"steps": 1,
"model": [
"1",
0
]
},
"class_type": "SDTurboScheduler"
},
"10": {
"inputs": {
"width": 512,
"height": 512,
"batch_size": 1
},
"class_type": "EmptyLatentImage"
},
"11": {
"inputs": {
"samples": [
"4",
0
],
"vae": [
"1",
2
]
},
"class_type": "VAEDecode"
},
"12": {
"inputs": {
"images": [
"11",
0
]
},
"class_type": "PreviewImage"
},
"13": {
"inputs": {
"sampler_name": "euler"
},
"class_type": "KSamplerSelect"
},
"19": {
"inputs": {
"pixels": [
"21",
0
],
"vae": [
"1",
2
]
},
"class_type": "VAEEncode"
},
"21": {
"inputs": {
"upscale_method": "nearest-exact",
"scale_by": 2,
"image": [
"11",
0
]
},
"class_type": "ImageScaleBy"
},
"25": {
"inputs": {
"samples": [
"27",
0
],
"vae": [
"1",
2
]
},
"class_type": "VAEDecode"
},
"26": {
"inputs": {
"images": [
"25",
0
]
},
"class_type": "PreviewImage"
},
"27": {
"inputs": {
"seed": 340269611110554,
"steps": 2,
"cfg": 1,
"sampler_name": "ddim",
"scheduler": "ddim_uniform",
"denoise": 1,
"model": [
"1",
0
],
"positive": [
"2",
0
],
"negative": [
"3",
0
],
"latent_image": [
"19",
0
]
},
"class_type": "KSampler"
},
"28": {
"inputs": {
"filename_prefix": "ComfyUI",
"images": [
"25",
0
]
},
"class_type": "SaveImage"
},
"29": {
"inputs": {
"filename_prefix": "ComfyUI",
"images": [
"11",
0
]
},
"class_type": "SaveImage"
}
}