本文基于ComfyUI API编写了类似于webUI的Gradio交互式界面，支持文生图/图生图（SD1.x，SD2.x，SDXL，Stable Cascade），Lora，ControlNet，图生视频（SVD_xt_1_1），图像修复（FaceDetailer），图像放大（Extras），图片/API工作流信息读取（Info）。
1. 在线体验
本文代码已部署到百度飞桨AI Studio平台，以供大家在线体验Stable Diffusion ComfyUI/webUI 原版界面及自制Gradio界面。
项目链接：Stable Diffusion ComfyUI 在线体验
2. 自制Gradio界面展示
Stable Diffusion 文生图/图生图界面：
Stable Cascade 文生图/图生图界面：
Stable Video Diffusion 图生视频界面：
图片放大界面：
图片/API工作流信息读取界面：
3. Gradio界面设计及ComfyUI API调用源码
import io
import json
import os
import random
import re
import subprocess
import sys
import urllib.parse
import uuid
sys.path.append("/home/aistudio/ComfyUI/venv/lib/python3.10/site-packages")
import gradio as gr
import requests
import websocket
from collections import OrderedDict, deque
from PIL import Image
class Default:
# 1表示启用，0表示禁用
design_mode = 0
lora_weight = 0.8
controlnet_num = 5
controlnet_saveimage = 1
facedetailer_num = 3
prompt = "(best quality:1), (high quality:1), detailed/(extreme, highly, ultra/), realistic, 1girl/(beautiful, delicate, perfect/), "
negative_prompt = "(worst quality:1), (low quality:1), (normal quality:1), lowres, signature, blurry, watermark, duplicate, bad link, plump, bad anatomy, extra arms, extra digits, missing finger, bad hands, bad feet, deformed, error, mutation, text"
if design_mode == 1:
width = 64
hight = 64
steps = 2
else:
width = 512
hight = 768
steps = 20
class Initial:
os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
client_id = str(uuid.uuid4())
server_address = "127.0.0.1:8188"
if Default.design_mode == 0:
cmd = "ps -eo pid,args | grep 'export GRADIO_SERVER_PORT=' | awk '{print $8, $14}'"
ps_output = subprocess.run(cmd, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True).stdout.splitlines()
for i in ps_output:
if "/home/aistudio/ComfyUI.gradio.py" in i:
port = i.split(" ")[0].split("=")[1]
server_address = f"127.0.0.1:{port}"
output_dir = os.path.join(os.getcwd(), "ComfyUI/output/")
uploaded_image = {}
class Choices:
ws = websocket.WebSocket()
ws.connect("ws://{}/ws?clientId={}".format(Initial.server_address, Initial.client_id))
object_info = requests.get(url="http://{}/object_info".format(Initial.server_address)).json()
embedding = requests.get(url="http://{}/embeddings".format(Initial.server_address)).json()
ws.close()
ckpt = []
ckpt_list = {}
ckpt_name = object_info["ImageOnlyCheckpointLoader"]["input"]["required"]["ckpt_name"][0]
hidden_ckpt = ["stable_cascade_stage_c.safetensors", "stable_cascade_stage_b.safetensors", "svd_xt_1_1.safetensors", "control_v11p_sd15_canny.safetensors", "control_v11f1p_sd15_depth.safetensors", "control_v11p_sd15_openpose.safetensors"]
for i in ckpt_name:
path, file = os.path.split(i)
if file not in hidden_ckpt:
ckpt.append(file)
ckpt_list[file] = i
ckpt = sorted(ckpt)
controlnet_model = []
controlnet_model_list = {}
controlnet_name = object_info["ControlNetLoader"]["input"]["required"]["control_net_name"][0]
for i in controlnet_name:
path, file = os.path.split(i)
controlnet_model.append(file)
controlnet_model_list[file] = i
controlnet_model = sorted(controlnet_model)
preprocessor = ["Canny"]
if "AIO_Preprocessor" in object_info:
preprocessor = ["none", "Canny", "CannyEdgePreprocessor", "DepthAnythingPreprocessor", "DWPreprocessor", "OpenposePreprocessor"]
for i in sorted(object_info["AIO_Preprocessor"]["input"]["optional"]["preprocessor"][0]):
if i not in preprocessor:
preprocessor.append(i)
if "FaceDetailer" in object_info:
facedetailer_detector_model = []
facedetailer_detector_model_list = {}
facedetailer_detector_model_name = object_info["UltralyticsDetectorProvider"]["input"]["required"]["model_name"][0]
for i in facedetailer_detector_model_name:
path, file = os.path.split(i)
facedetailer_detector_model.append(file)
facedetailer_detector_model_list[file] = i
facedetailer_detector_model = sorted(facedetailer_detector_model)
lora = object_info["LoraLoader"]["input"]["required"]["lora_name"][0]
sampler = object_info["KSampler"]["input"]["required"]["sampler_name"][0]
scheduler = object_info["KSampler"]["input"]["required"]["scheduler"][0]
upscale_method = object_info["ImageScaleBy"]["input"]["required"]["upscale_method"][0]
upscale_model = object_info["UpscaleModelLoader"]["input"]["required"]["model_name"][0]
vae = ["Automatic"]
for i in sorted(object_info["VAELoader"]["input"]["required"]["vae_name"][0]):
vae.append(i)
class Function:
def format_prompt(prompt):
prompt = re.sub(r"\s+,", ",", prompt)
prompt = re.sub(r"\s+", " ", prompt)
prompt = re.sub(",,+", ",", prompt)
prompt = re.sub(",", ", ", prompt)
prompt = re.sub(r"\s+", " ", prompt)
prompt = re.sub(r"^,", "", prompt)
prompt = re.sub(r"^ ", "", prompt)
prompt = re.sub(r" $", "", prompt)
prompt = re.sub(r",$", "", prompt)
prompt = re.sub(": ", ":", prompt)
return prompt
def get_model_path(model_name):
return Choices.ckpt_list[model_name]
def gen_seed(seed):
seed = int(seed)
if seed < 0:
seed = random.randint(0, 18446744073709551615)
if seed > 18446744073709551615:
seed = 18446744073709551615
return seed
def initialize():
Lora.cache = {}
Upscale.cache = {}
UpscaleWithModel.cache = {}
ControlNet.cache = {}
FaceDetailer.cache = {}
def upload_image(image):
buffer = io.BytesIO()
image.save(buffer, format="png")
image = buffer.getbuffer()
image_hash = hash(image.tobytes())
if image_hash in Initial.uploaded_image:
return Initial.uploaded_image[image_hash]
image_name = str(uuid.uuid4()) + ".png"
Initial.uploaded_image[image_hash] = image_name
image_file = {"image": (image_name, image)}
ws = websocket.WebSocket()
ws.connect("ws://{}/ws?clientId={}".format(Initial.server_address, Initial.client_id))
requests.post(url="http://{}/upload/image".format(Initial.server_address), files=image_file)
ws.close()
return image_name
def order_workflow(workflow):
link_list = {}
for node in workflow:
node_link = []
for input in workflow[node]["inputs"]:
if isinstance(workflow[node]["inputs"][input], list):
node_link.append(workflow[node]["inputs"][input][0])
link_list[node] = node_link
in_degree = {v: 0 for v in link_list}
for node in link_list:
for neighbor in link_list[node]:
in_degree[neighbor] += 1
queue = deque([node for node in in_degree if in_degree[node] == 0])
order_list = []
while queue:
node = queue.popleft()
order_list.append(node)
for neighbor in link_list[node]:
in_degree[neighbor] -= 1
if in_degree[neighbor] == 0:
queue.append(neighbor)
order_list = order_list[::-1]
max_nodes = 1000
new_node_id = max_nodes * 10 + 1
workflow_string = json.dumps(workflow)
for node in order_list:
workflow_string = workflow_string.replace(f'"{node}"', f'"{new_node_id}"')
new_node_id += 1
workflow = json.loads(workflow_string)
workflow = OrderedDict(sorted(workflow.items()))
new_node_id = 1
workflow_string = json.dumps(workflow)
for node in workflow:
workflow_string = workflow_string.replace(f'"{node}"', f'"{new_node_id}"')
new_node_id += 1
workflow = json.loads(workflow_string)
for node in workflow:
if "_meta" in workflow[node]:
del workflow[node]["_meta"]
return workflow
def post_interrupt():
Initial.interrupt = True
ws = websocket.WebSocket()
ws.connect("ws://{}/ws?clientId={}".format(Initial.server_address, Initial.client_id))
requests.post(url="http://{}/interrupt".format(Initial.server_address))
ws.close()
def add_embedding(embedding, negative_prompt):
for i in Choices.embedding:
negative_prompt = negative_prompt.replace(f"embedding:{i},", "")
negative_prompt = Function.format_prompt(negative_prompt)
for i in embedding[::-1]:
negative_prompt = f"embedding:{i}, {negative_prompt}"
return negative_prompt
def gen_image(workflow, counter, batch_count, progress):
if counter == 1:
progress(0, desc="Processing...")
if batch_count == 1:
batch_info = ""
else:
batch_info = f"Batch {counter}/{batch_count}: "
workflow = Function.order_workflow(workflow)
current_progress = 0
Initial.interrupt = False
ws = websocket.WebSocket()
ws.connect("ws://{}/ws?clientId={}".format(Initial.server_address, Initial.client_id))
data = {"prompt": workflow, "client_id": Initial.client_id}
prompt_id = requests.post(url="http://{}/prompt".format(Initial.server_address), json=data).json()["prompt_id"]
while True:
try:
ws.settimeout(0.1)
wsrecv = ws.recv()
if isinstance(wsrecv, str):
data = json.loads(wsrecv)["data"]
if "node" in data:
if data["node"] is not None:
if "value" in data and "max" in data:
if data["max"] > 1:
current_progress = data["value"]/data["max"]
progress(current_progress, desc=f"{batch_info}" + workflow[data["node"]]["class_type"] + " " + str(data["value"]) + "/" + str(data["max"]))
else:
progress(current_progress, desc=f"{batch_info}" + workflow[data["node"]]["class_type"])
if data["node"] is None and data["prompt_id"] == prompt_id:
break
else:
continue
except websocket.WebSocketTimeoutException:
if Initial.interrupt is True:
ws.close()
return None, None
history = requests.get(url="http://{}/history/{}".format(Initial.server_address, prompt_id)).json()[prompt_id]
images = []
file_path = ""
for node_id in history["outputs"]:
node_output = history["outputs"][node_id]
if "images" in node_output:
for image in node_output["images"]:
file_path = Initial.output_dir + image["filename"]
data = {"filename": image["filename"], "subfolder": image["subfolder"], "type": image["type"]}
url_values = urllib.parse.urlencode(data)
image_data = requests.get("http://{}/view?{}".format(Initial.server_address, url_values)).content
image = Image.open(io.BytesIO(image_data))
images.append(image)
ws.close()
return images, file_path
def get_gallery_index(evt: gr.SelectData):
return evt.index
def get_image_info(image_pil):
image_info=[]
if image_pil is None:
return
for key, value in image_pil.info.items():
image_info.append(value)
if image_info != []:
image_info = image_info[0]
if image_info == 0:
image_info = "None"
else:
image_info = "None"
return image_info
def send_to(data, index):
if data == [] or data is None:
return None
return data[index]
class Lora:
cache = {}
def add_node(module, workflow, node_id, model_port, clip_port):
for lora in Lora.cache[module]:
strength_model = Lora.cache[module][lora]
strength_clip = Lora.cache[module][lora]
node_id += 1
workflow[str(node_id)] = {"inputs": {"lora_name": lora, "strength_model": strength_model, "strength_clip": strength_clip, "model": model_port, "clip": clip_port}, "class_type": "LoraLoader"}
model_port = [str(node_id), 0]
clip_port = [str(node_id), 1]
return workflow, node_id, model_port, clip_port
def update_cache(module, lora, lora_weight):
if Initial.initialized is False:
Function.initialize()
if lora == []:
Lora.cache[module] = {}
return True, [], gr.update(value="", visible=False)
lora_list = {}
for i in lora_weight.split("<"):
for j in i.split(">"):
if j != "" and ":" in j:
lora_name, weight = j.split(":")
lora_list[lora_name] = weight
lora_weight = ""
Lora.cache[module] = {}
for i in lora:
if i in lora_list:
weight = lora_list[i]
else:
weight = Default.lora_weight
if lora.index(i) == 0:
lora_weight = f"<{i}:{weight}>"
else:
lora_weight = f"{lora_weight}\n\n<{i}:{weight}>"
if weight != "":
weight = float(weight)
Lora.cache[module][i] = weight
return True, gr.update(), gr.update(value=lora_weight, visible=True)
def blocks(module):
module = gr.Textbox(value=module, visible=False)
lora = gr.Dropdown(Choices.lora, label="Lora", multiselect=True, interactive=True)
lora_weight = gr.Textbox(label="Lora weight | Lora 权重", visible=False)
for gr_block in [lora, lora_weight]:
gr_block.change(fn=Lora.update_cache, inputs=[module, lora, lora_weight], outputs=[Initial.initialized, lora, lora_weight])
class Upscale:
cache = {}
def add_node(module, workflow, node_id, image_port):
upscale_method = Upscale.cache[module]["upscale_method"]
scale_by = Upscale.cache[module]["scale_by"]
node_id += 1
workflow[str(node_id)] = {"inputs": {"upscale_method": upscale_method, "scale_by": scale_by, "image": image_port}, "class_type": "ImageScaleBy"}
image_port = [str(node_id), 0]
return workflow, node_id, image_port
def auto_enable(scale_by):
if scale_by > 1:
return True
else:
return False
def update_cache(module, enable, upscale_method, scale_by):
if Initial.initialized is False:
Function.initialize()
if module not in Upscale.cache:
Upscale.cache[module] = {}
if enable is True:
Upscale.cache[module]["upscale_method"] = upscale_method
Upscale.cache[module]["scale_by"] = scale_by
else:
del Upscale.cache[module]
return True
def blocks(module):
module = gr.Textbox(value=module, visible=False)
enable = gr.Checkbox(label="Enable（放大系数大于1后自动启用）")
with gr.Row():
upscale_method = gr.Dropdown(Choices.upscale_method, label="Upscale method | 放大方法", value=Choices.upscale_method[-1])
scale_by = gr.Slider(minimum=1, maximum=8, step=1, label="Scale by | 放大系数", value=1)
scale_by.release(fn=Upscale.auto_enable, inputs=[scale_by], outputs=[enable])
inputs = [module, enable, upscale_method, scale_by]
for gr_block in inputs:
if type(gr_block) is gr.components.slider.Slider:
gr_block.release(fn=Upscale.update_cache, inputs=inputs, outputs=[Initial.initialized])
else:
gr_block.change(fn=Upscale.update_cache, inputs=inputs, outputs=[Initial.initialized])
class UpscaleWithModel:
cache = {}
def add_node(module, workflow, node_id, image_port):
upscale_model = UpscaleWithModel.cache[module]["upscale_model"]
node_id += 1
workflow[str(node_id)] = {"inputs": {"model_name": upscale_model}, "class_type": "UpscaleModelLoader"}
upscale_model_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"upscale_model": upscale_model_port, "image": image_port}, "class_type": "ImageUpscaleWithModel"}
image_port = [str(node_id), 0]
return workflow, node_id, image_port
def update_cache(module, enable, upscale_model):
if Initial.initialized is False:
Function.initialize()
if module not in UpscaleWithModel.cache:
UpscaleWithModel.cache[module] = {}
if enable is True:
UpscaleWithModel.cache[module]["upscale_model"] = upscale_model
else:
del UpscaleWithModel.cache[module]
return True
def blocks(module):
module = gr.Textbox(value=module, visible=False)
enable = gr.Checkbox(label="Enable")
upscale_model = gr.Dropdown(Choices.upscale_model, label="Upscale model | 超分模型", value=Choices.upscale_model[0])
inputs = [module, enable, upscale_model]
for gr_block in inputs:
gr_block.change(fn=UpscaleWithModel.update_cache, inputs=inputs, outputs=[Initial.initialized])
class ControlNet:
cache = {}
model_preprocessor_list = {
"control_v11e_sd15_ip2p.safetensors": [],
"control_v11e_sd15_shuffle.safetensors": ["ShufflePreprocessor"],
"control_v11f1e_sd15_tile.bin": ["TilePreprocessor", "TTPlanet_TileGF_Preprocessor", "TTPlanet_TileSimple_Preprocessor"],
"control_v11f1p_sd15_depth.safetensors": ["DepthAnythingPreprocessor", "LeReS-DepthMapPreprocessor", "MiDaS-NormalMapPreprocessor", "MeshGraphormer-DepthMapPreprocessor", "MeshGraphormer+ImpactDetector-DepthMapPreprocessor", "MiDaS-DepthMapPreprocessor", "Zoe_DepthAnythingPreprocessor", "Zoe-DepthMapPreprocessor"],
"control_v11p_sd15_canny.safetensors": ["Canny", "CannyEdgePreprocessor"],
"control_v11p_sd15_inpaint.safetensors": [],
"control_v11p_sd15_lineart.safetensors": ["LineArtPreprocessor", "LineartStandardPreprocessor"],
"control_v11p_sd15_mlsd.safetensors": ["M-LSDPreprocessor"],
"control_v11p_sd15_normalbae.safetensors": ["BAE-NormalMapPreprocessor", "DSINE-NormalMapPreprocessor"],
"control_v11p_sd15_openpose.safetensors": ["DWPreprocessor", "OpenposePreprocessor", "DensePosePreprocessor"],
"control_v11p_sd15_scribble.safetensors": ["ScribblePreprocessor", "Scribble_XDoG_Preprocessor", "Scribble_PiDiNet_Preprocessor", "FakeScribblePreprocessor"],
"control_v11p_sd15_seg.safetensors": ["AnimeFace_SemSegPreprocessor", "OneFormer-COCO-SemSegPreprocessor", "OneFormer-ADE20K-SemSegPreprocessor", "SemSegPreprocessor", "UniFormer-SemSegPreprocessor"],
"control_v11p_sd15_softedge.safetensors": ["HEDPreprocessor", "PiDiNetPreprocessor", "TEEDPreprocessor", "DiffusionEdge_Preprocessor"],
"control_v11p_sd15s2_lineart_anime.safetensors": ["AnimeLineArtPreprocessor", "Manga2Anime_LineArt_Preprocessor"],
"control_scribble.safetensors": ["BinaryPreprocessor"],
"ioclab_sd15_recolor.safetensors": ["ImageLuminanceDetector", "ImageIntensityDetector"],
"control_sd15_animal_openpose_fp16.pth": ["AnimalPosePreprocessor"],
"controlnet_sd21_laion_face_v2.safetensors": ["MediaPipe-FaceMeshPreprocessor"]
}
def add_node(module, counter, workflow, node_id, positive_port, negative_port):
for unit_id in ControlNet.cache[module]:
preprocessor = ControlNet.cache[module][unit_id]["preprocessor"]
model = ControlNet.cache[module][unit_id]["model"]
input_image = ControlNet.cache[module][unit_id]["input_image"]
resolution = ControlNet.cache[module][unit_id]["resolution"]
strength = ControlNet.cache[module][unit_id]["strength"]
start_percent = ControlNet.cache[module][unit_id]["start_percent"]
end_percent = ControlNet.cache[module][unit_id]["end_percent"]
node_id += 1
workflow[str(node_id)] = {"inputs": {"image": input_image, "upload": "image"}, "class_type": "LoadImage"}
image_port = [str(node_id), 0]
if preprocessor == "Canny":
node_id += 1
workflow[str(node_id)] = {"inputs": {"low_threshold": 0.3, "high_threshold": 0.7, "image": image_port}, "class_type": "Canny"}
image_port = [str(node_id), 0]
else:
node_id += 1
workflow[str(node_id)] = {"inputs": {"preprocessor": preprocessor, "resolution": resolution, "image": image_port}, "class_type": "AIO_Preprocessor"}
image_port = [str(node_id), 0]
if counter == 1 and Default.controlnet_saveimage == 1:
node_id += 1
workflow[str(node_id)] = {"inputs": {"filename_prefix": "ControlNet", "images": image_port}, "class_type": "SaveImage"}
node_id += 1
workflow[str(node_id)] = {"inputs": {"control_net_name": model}, "class_type": "ControlNetLoader"}
control_net_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"strength": strength, "start_percent": start_percent, "end_percent": end_percent, "positive": positive_port, "negative": negative_port, "control_net": control_net_port, "image": image_port}, "class_type": "ControlNetApplyAdvanced"}
positive_port = [str(node_id), 0]
negative_port = [str(node_id), 1]
return workflow, node_id, positive_port, negative_port
def auto_enable():
return True
def auto_select_model(preprocessor):
for model in Choices.controlnet_model:
if model in ControlNet.model_preprocessor_list:
if preprocessor in ControlNet.model_preprocessor_list[model]:
return gr.update(value=model)
return gr.update(value="未定义/检测到对应的模型，请自行选择！")
def preprocess(unit_id, preview, preprocessor, input_image, resolution, progress=gr.Progress()):
if preview is False or input_image is None:
return
input_image = Function.upload_image(input_image)
workflow = {}
node_id = 1
workflow[str(node_id)] = {"inputs": {"image": input_image, "upload": "image"}, "class_type": "LoadImage"}
image_port = [str(node_id), 0]
if preprocessor == "Canny":
node_id += 1
workflow[str(node_id)] = {"inputs": {"low_threshold": 0.3, "high_threshold": 0.7, "image": image_port}, "class_type": "Canny"}
image_port = [str(node_id), 0]
else:
node_id += 1
workflow[str(node_id)] = {"inputs": {"preprocessor": preprocessor, "resolution": resolution, "image": image_port}, "class_type": "AIO_Preprocessor"}
image_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"images": image_port}, "class_type": "PreviewImage"}
output = Function.gen_image(workflow, 1, 1, progress)[0]
if output is not None:
output = output[0]
return output
def update_cache(module, unit_id, enable, preprocessor, model, input_image, resolution, strength, start_percent, end_percent):
if Initial.initialized is False:
Function.initialize()
if module not in ControlNet.cache:
ControlNet.cache[module] = {}
ControlNet.cache[module][unit_id] = {}
if input_image is None:
del ControlNet.cache[module][unit_id]
return True, False
if model not in Choices.controlnet_model:
del ControlNet.cache[module][unit_id]
return True, False
if enable is True:
ControlNet.cache[module][unit_id]["preprocessor"] = preprocessor
ControlNet.cache[module][unit_id]["model"] = Choices.controlnet_model_list[model]
ControlNet.cache[module][unit_id]["input_image"] = Function.upload_image(input_image)
ControlNet.cache[module][unit_id]["resolution"] = resolution
ControlNet.cache[module][unit_id]["strength"] = strength
ControlNet.cache[module][unit_id]["start_percent"] = start_percent
ControlNet.cache[module][unit_id]["end_percent"] = end_percent
else:
del ControlNet.cache[module][unit_id]
return True, gr.update()
def unit(module, i):
module = gr.Textbox(value=module, visible=False)
unit_id = gr.Textbox(value=i, visible=False)
with gr.Row():
enable = gr.Checkbox(label="Enable（上传图片后自动启用）")
preview = gr.Checkbox(label="Preview")
with gr.Row():
preprocessor = gr.Dropdown(Choices.preprocessor, label="Preprocessor", value="Canny")
model = gr.Dropdown(Choices.controlnet_model, label="ControlNet model", value="control_v11p_sd15_canny.safetensors")
with gr.Row():
input_image = gr.Image(type="pil")
preprocess_preview = gr.Image(label="Preprocessor preview")
with gr.Row():
resolution = gr.Slider(label="Resolution", minimum=64, maximum=2048, step=64, value=512)
strength = gr.Slider(label="Strength", minimum=0, maximum=2, step=0.01, value=1)
with gr.Row():
start_percent = gr.Slider(label="Start percent", minimum=0, maximum=1, step=0.01, value=0)
end_percent = gr.Slider(label="End percent", minimum=0, maximum=1, step=0.01, value=1)
input_image.upload(fn=ControlNet.auto_enable, inputs=None, outputs=[enable])
preprocessor.change(fn=ControlNet.auto_select_model, inputs=[preprocessor], outputs=[model])
for gr_block in [preview, preprocessor, input_image]:
gr_block.change(fn=ControlNet.preprocess, inputs=[unit_id, preview, preprocessor, input_image, resolution], outputs=[preprocess_preview])
inputs = [module, unit_id, enable, preprocessor, model, input_image, resolution, strength, start_percent, end_percent]
for gr_block in inputs:
if type(gr_block) is gr.components.slider.Slider:
gr_block.release(fn=ControlNet.update_cache, inputs=inputs, outputs=[Initial.initialized, enable])
else:
gr_block.change(fn=ControlNet.update_cache, inputs=inputs, outputs=[Initial.initialized, enable])
def blocks(module):
with gr.Tab(label="控制网络"):
if Default.controlnet_num == 1:
ControlNet.unit(module, 1)
else:
for i in range(Default.controlnet_num):
with gr.Tab(label=f"ControlNet Unit {i + 1}"):
ControlNet.unit(module, i + 1)
class FaceDetailer:
cache = {}
def add_node(module, workflow, node_id, image_port, model_port, clip_port, vae_port, positive_port, negative_port, seed, steps, cfg, sampler_name, scheduler):
for unit_id in FaceDetailer.cache[module]:
model = Choices.facedetailer_detector_model_list[FaceDetailer.cache[module][unit_id]["model"]]
node_id += 1
workflow[str(node_id)] = {"inputs": {"model_name": model}, "class_type": "UltralyticsDetectorProvider"}
bbox_detector_port = [str(node_id), 0]
segm_detector_opt_port = [str(node_id), 1]
node_id += 1
workflow[str(node_id)] = {"inputs": {"model_name": "sam_vit_b_01ec64.pth", "device_mode": "AUTO"}, "class_type": "SAMLoader"}
sam_model_opt_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"guide_size": 384, "guide_size_for": "True", "max_size": 1024, "seed": seed, "steps": steps, "cfg": cfg, "sampler_name": sampler_name, "scheduler": scheduler, "denoise": 0.5, "feather": 5, "noise_mask": "True", "force_inpaint": "True", "bbox_threshold": 0.5, "bbox_dilation": 10, "bbox_crop_factor": 3, "sam_detection_hint": "center-1", "sam_dilation": 0, "sam_threshold": 0.93, "sam_bbox_expansion": 0, "sam_mask_hint_threshold": 0.7, "sam_mask_hint_use_negative": "False", "drop_size": 10, "wildcard": "", "cycle": 1, "inpaint_model": "False", "noise_mask_feather": 20, "image": image_port, "model": model_port, "clip": clip_port, "vae": vae_port, "positive": positive_port, "negative": negative_port, "bbox_detector": bbox_detector_port, "sam_model_opt": sam_model_opt_port, "segm_detector_opt": segm_detector_opt_port}, "class_type": "FaceDetailer"}
image_port = [str(node_id), 0]
return workflow, node_id, image_port
def update_cache(module, unit_id, enable, model):
if Initial.initialized is False:
Function.initialize()
if module not in FaceDetailer.cache:
FaceDetailer.cache[module] = {}
FaceDetailer.cache[module][unit_id] = {}
if enable is True:
FaceDetailer.cache[module][unit_id]["model"] = model
else:
del FaceDetailer.cache[module][unit_id]
return True
def unit(module, i):
module = gr.Textbox(value=module, visible=False)
unit_id = gr.Textbox(value=i, visible=False)
enable = gr.Checkbox(label="Enable")
if i == 1:
model = gr.Dropdown(Choices.facedetailer_detector_model, label="Detector model", value="face_yolov8m.pt")
if i == 2:
model = gr.Dropdown(Choices.facedetailer_detector_model, label="Detector model", value="hand_yolov8s.pt")
if i == 3:
model = gr.Dropdown(Choices.facedetailer_detector_model, label="Detector model", value="person_yolov8m-seg.pt")
inputs = [module, unit_id, enable, model]
for gr_block in inputs:
gr_block.change(fn=FaceDetailer.update_cache, inputs=inputs, outputs=[Initial.initialized])
def blocks(module):
with gr.Tab(label="图像修复"):
if Default.facedetailer_num == 1:
FaceDetailer.unit(module, 1)
else:
with gr.Row():
for i in range(Default.facedetailer_num):
with gr.Column():
with gr.Tab(label=f"FaceDetailer Unit {i + 1}"):
FaceDetailer.unit(module, i + 1)
if Default.facedetailer_num % 2 != 0:
with gr.Column():
gr.HTML("")
class Postprocess:
def add_node(module, *args):
if module == "SD":
workflow, node_id, image_port, model_port, clip_port, vae_port, positive_port, negative_port, seed, steps, cfg, sampler_name, scheduler = args
else:
workflow, node_id, image_port = args
if module in FaceDetailer.cache:
workflow, node_id, image_port = FaceDetailer.add_node(module, workflow, node_id, image_port, model_port, clip_port, vae_port, positive_port, negative_port, seed, steps, cfg, sampler_name, scheduler)
if module in Upscale.cache:
workflow, node_id, image_port = Upscale.add_node(module, workflow, node_id, image_port)
if module in UpscaleWithModel.cache:
workflow, node_id, image_port = UpscaleWithModel.add_node(module, workflow, node_id, image_port)
return workflow, node_id, image_port
def blocks(module):
if module == "SD":
if "FaceDetailer" in Choices.object_info:
FaceDetailer.blocks(module)
with gr.Tab(label="图像放大"):
with gr.Row():
with gr.Tab(label="算术放大"):
Upscale.blocks(module)
with gr.Row():
with gr.Tab(label="超分放大"):
UpscaleWithModel.blocks(module)
gr.HTML("注意：同时启用两种放大模式将先执行算术放大，再执行超分放大，最终放大倍数为二者放大倍数的乘积！")
class SD:
def generate(initialized, batch_count, ckpt_name, vae_name, clip_mode, clip_skip, width, height, batch_size, negative_prompt, positive_prompt, seed, steps, cfg, sampler_name, scheduler, denoise, input_image, progress=gr.Progress()):
module = "SD"
ckpt_name = Function.get_model_path(ckpt_name)
seed = Function.gen_seed(seed)
if input_image is not None:
input_image = Function.upload_image(input_image)
counter = 1
output_images = []
node_id = 0
while counter <= batch_count:
workflow = {}
node_id += 1
workflow[str(node_id)] = {"inputs": {"ckpt_name": ckpt_name}, "class_type": "CheckpointLoaderSimple"}
model_port = [str(node_id), 0]
clip_port = [str(node_id), 1]
vae_port = [str(node_id), 2]
if vae_name != "Automatic":
node_id += 1
workflow[str(node_id)] = {"inputs": {"vae_name": vae_name}, "class_type": "VAELoader"}
vae_port = [str(node_id), 0]
if input_image is None:
node_id += 1
workflow[str(node_id)] = {"inputs": {"width": width, "height": height, "batch_size": batch_size}, "class_type": "EmptyLatentImage"}
latent_image_port = [str(node_id), 0]
else:
node_id += 1
workflow[str(node_id)] = {"inputs": {"image": input_image, "upload": "image"}, "class_type": "LoadImage"}
pixels_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"pixels": pixels_port, "vae": vae_port}, "class_type": "VAEEncode"}
latent_image_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"stop_at_clip_layer": -clip_skip, "clip": clip_port}, "class_type": "CLIPSetLastLayer"}
clip_port = [str(node_id), 0]
if initialized is True and module in Lora.cache:
workflow, node_id, model_port, clip_port = Lora.add_node(module, workflow, node_id, model_port, clip_port)
node_id += 1
if clip_mode == "ComfyUI":
workflow[str(node_id)] = {"inputs": {"text": positive_prompt, "clip": clip_port}, "class_type": "CLIPTextEncode"}
else:
workflow[str(node_id)] = {"inputs": {"text": positive_prompt, "token_normalization": "none", "weight_interpretation": "A1111", "clip": clip_port}, "class_type": "BNK_CLIPTextEncodeAdvanced"}
positive_port = [str(node_id), 0]
node_id += 1
if clip_mode == "ComfyUI":
workflow[str(node_id)] = {"inputs": {"text": negative_prompt, "clip": clip_port}, "class_type": "CLIPTextEncode"}
else:
workflow[str(node_id)] = {"inputs": {"text": negative_prompt, "token_normalization": "none", "weight_interpretation": "A1111", "clip": clip_port}, "class_type": "BNK_CLIPTextEncodeAdvanced"}
negative_port = [str(node_id), 0]
if initialized is True and module in ControlNet.cache:
workflow, node_id, positive_port, negative_port = ControlNet.add_node(module, counter, workflow, node_id, positive_port, negative_port)
node_id += 1
workflow[str(node_id)] = {"inputs": {"seed": seed, "steps": steps, "cfg": cfg, "sampler_name": sampler_name, "scheduler": scheduler, "denoise": denoise, "model": model_port, "positive": positive_port, "negative": negative_port, "latent_image": latent_image_port}, "class_type": "KSampler"}
samples_port = [str(node_id), 0]
node_id += 1
workflow[str(node_id)] = {"inputs": {"samples": samples_port, "vae": vae_port}, "class_type": "VAEDecode"}
image_port = [str(node_id), 0]
if initialized is True:
workflow, node_id, image_port = Postprocess.add_node(module, workflow, node_id, image_port, model_port, clip_port, vae_port, positive_port, negative_port, seed, steps, cfg, sampler_name, scheduler)
node_id += 1
workflow[str(node_id)] = {"inputs": {"filename_prefix": "ComfyUI", "images": image_port}, "class_type": "SaveImage"}
images = Function.gen_image(workflow, counter, batch_count, progress)[0]
if images is None:
break
for image in images:
output_images.append(image)
seed += 1
counter += 1
return output_images, output_images
def blocks():
with gr.Row():
with gr.Column():
positive_prompt = gr.Textbox(placeholder="Positive prompt | 正向提示词", show_label=False, value=Default.prompt, lines=3)
negative_prompt = gr.Textbox(placeholder="Negative prompt | 负向提示词", show_label=False, value=Default.negative_prompt, lines=3)
with gr.Tab(label="基础设置"):
with gr.Row():
ckpt_name = gr.Dropdown(Choices.ckpt, label="Ckpt name | Ckpt 模型名称", value=Choices.ckpt[0])
vae_name = gr.Dropdown(Choices.vae, label="VAE name | VAE 模型名称", value=Choices.vae[0])
if "BNK_CLIPTextEncodeAdvanced" in Choices.object_info:
clip_mode = gr.Dropdown(["ComfyUI", "WebUI"], label="Clip 编码类型", value="ComfyUI")
else:
clip_mode = gr.Dropdown(["ComfyUI", "WebUI"], label="Clip 编码类型", value="ComfyUI", visible=False)
clip_skip = gr.Slider(minimum=1, maximum=12, step=1, label="Clip 跳过", value=1)
with gr.Row():
width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width | 图像宽度", value=Default.width)
batch_size = gr.Slider(minimum=1, maximum=8, step=1, label="Batch size | 批次大小", value=1)
with gr.Row():
height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height | 图像高度", value=Default.hight)
batch_count = gr.Slider(minimum=1, maximum=100, step=1, label="Batch count | 生成批次", value=1)
with gr.Row():
if Choices.lora != []:
Lora.blocks("SD")
if Choices.embedding != []:
embedding = gr.Dropdown(Choices.embedding, label="Embedding", multiselect=True, interactive=True)
embedding.change(fn=Function.add_embedding, inputs=[embedding, negative_prompt], outputs=[negative_prompt])
with gr.Row():
SD.input_image = gr.Image(value=None, type="pil")
gr.HTML("<br>上传图片即自动转为图生图模式。<br><br>文生图、图生图模式共享设置参数。<br><br>图像宽度、图像高度、批次大小对图生图无效。")
with gr.Tab(label="采样设置"):
with gr.Row():
sampler_name = gr.Dropdown(Choices.sampler, label="Sampling method | 采样方法", value=Choices.sampler[12])
scheduler = gr.Dropdown(Choices.scheduler, label="Schedule type | 采样计划表类型", value=Choices.scheduler[1])
with gr.Row():
denoise = gr.Slider(minimum=0, maximum=1, step=0.05, label="Denoise | 去噪强度", value=1)
steps = gr.Slider(minimum=1, maximum=100, step=1, label="Sampling steps | 采样次数", value=Default.steps)
with gr.Row():
cfg = gr.Slider(minimum=0, maximum=20, step=0.1, label="CFG Scale | CFG权重", value=7)
seed = gr.Slider(minimum=-1, maximum=18446744073709550000, step=1, label="Seed | 种子数", value=-1)
if Choices.controlnet_model != []:
ControlNet.blocks("SD")
Postprocess.blocks("SD")
with gr.Column():
with gr.Row():
btn = gr.Button("Generate | 生成", elem_id="button")
btn2 = gr.Button("Interrupt | 终止")
output = gr.Gallery(preview=True, height=600)
with gr.Row():
SD.send_to_sd = gr.Button("发送图片至 SD")
if SC.enable is True:
SD.send_to_sc = gr.Button("发送图片至 SC")
if SVD.enable is True:
SD.send_to_svd = gr.Button("发送图片至 SVD")
SD.send_to_extras = gr.Button("发送图片至 Extras")
SD.send_to_info = gr.Button("发送图片至 Info")
SD.data = gr.State()
SD.index = gr.State()
btn.click(fn=SD.generate, inputs=[Initial.initialized, batch_count, ckpt_name, vae_name, clip_mode, clip_skip, width, height, batch_size, negative_prompt, positive_prompt, seed, steps, cfg, sampler_name, scheduler, denoise, SD.input_image], outputs=[output, SD.data])
btn2.click(fn=Function.post_interrupt, inputs=None, outputs=None)
output.select(fn=Function.get_gallery_index, inputs=None, outputs=[SD.index])
class SC:
if Default.design_mode == 1:
enable = True
elif "stable_cascade_stage_c.safetensors" in Choices.ckpt_list and "stable_cascade_stage_b.safetensors" in Choices.ckpt_list:
enable = True
else:
enable = False
def generate(initialized, batch_count, positive_prompt, negative_prompt, width, height, batch_size, seed_c, steps_c, cfg_c, sampler_name_c, scheduler_c, denoise_c, seed_b, steps_b, cfg_b, sampler_name_b, scheduler_b, denoise_b, input_image, progress=gr.Progress()):
module = "SC"
ckpt_name_c = Function.get_model_path("stable_cascade_stage_c.safetensors")
ckpt_name_b = Function.get_model_path("stable_cascade_stage_b.safetensors")
seed_c = Function.gen_seed(seed_c)
seed_b = Function.gen_seed(seed_b)
if input_image is not None:
input_image = Function.upload_image(input_image)
counter = 1
output_images = []
while counter <= batch_count:
workflow = {
"1": {"inputs": {"ckpt_name": ckpt_name_c}, "class_type": "CheckpointLoaderSimple"},
"2": {"inputs": {"image": input_image, "upload": "image"}, "class_type": "LoadImage"},
"3": {"inputs": {"compression": 42, "image": ["2", 0], "vae": ["1", 2]}, "class_type": "StableCascade_StageC_VAEEncode"},
"4": {"inputs": {"text": negative_prompt, "clip": ["1", 1]}, "class_type": "CLIPTextEncode"},
"5": {"inputs": {"text": positive_prompt, "clip": ["1", 1]}, "class_type": "CLIPTextEncode"},
"6": {"inputs": {"seed": seed_c, "steps": steps_c, "cfg": cfg_c, "sampler_name": sampler_name_c, "scheduler": scheduler_c, "denoise": denoise_c, "model": ["1", 0], "positive": ["5", 0], "negative": ["4", 0], "latent_image": ["3", 0]}, "class_type": "KSampler"},
"7": {"inputs": {"conditioning": ["5", 0], "stage_c": ["6", 0]}, "class_type": "StableCascade_StageB_Conditioning"},
"8": {"inputs": {"ckpt_name": ckpt_name_b}, "class_type": "CheckpointLoaderSimple"},
"9": {"inputs": {"seed": seed_b, "steps": steps_b, "cfg": cfg_b, "sampler_name": sampler_name_b, "scheduler": scheduler_b, "denoise": denoise_b, "model": ["8", 0], "positive": ["7", 0], "negative": ["4", 0], "latent_image": ["3", 1]}, "class_type": "KSampler"},
"10": {"inputs": {"samples": ["9", 0], "vae": ["8", 2]}, "class_type": "VAEDecode"}
}
if input_image is None:
del workflow["2"]
workflow["3"] = {"inputs": {"width": width, "height": height, "compression": 42, "batch_size": batch_size}, "class_type": "StableCascade_EmptyLatentImage"}
node_id = 10
image_port = [str(node_id), 0]
if initialized is True:
workflow, node_id, image_port = Postprocess.add_node(module, workflow, node_id, image_port)
node_id += 1
workflow[str(node_id)] = {"inputs": {"filename_prefix": "ComfyUI", "images": image_port}, "class_type": "SaveImage"}
images = Function.gen_image(workflow, counter, batch_count, progress)[0]
if images is None:
break
for image in images:
output_images.append(image)
seed_c += 1
counter += 1
return output_images, output_images
def blocks():
with gr.Row():
with gr.Column():
positive_prompt = gr.Textbox(placeholder="Positive prompt | 正向提示词", show_label=False, value=Default.prompt, lines=3)
negative_prompt = gr.Textbox(placeholder="Negative prompt | 负向提示词", show_label=False, value=Default.negative_prompt, lines=3)
with gr.Tab(label="基础设置"):
with gr.Row():
width = gr.Slider(minimum=128, maximum=2048, step=128, label="Width | 图像宽度", value=1024)
batch_size = gr.Slider(minimum=1, maximum=8, step=1, label="Batch size | 批次大小", value=1)
with gr.Row():
height = gr.Slider(minimum=128, maximum=2048, step=128, label="Height | 图像高度", value=1024)
batch_count = gr.Slider(minimum=1, maximum=100, step=1, label="Batch count | 生成批次", value=1)
with gr.Row():
SC.input_image = gr.Image(value=None, type="pil")
gr.HTML("<br>上传图片即自动转为图生图模式。<br><br>文生图、图生图模式共享设置参数。<br><br>图像宽度、图像高度、批次大小对图生图无效。")
with gr.Tab(label="Stage C 采样设置"):
with gr.Row():
sampler_name_c = gr.Dropdown(Choices.sampler, label="Sampling method | 采样方法", value=Choices.sampler[12])
scheduler_c = gr.Dropdown(Choices.scheduler, label="Schedule type | 采样计划表类型", value=Choices.scheduler[1])
with gr.Row():
denoise_c = gr.Slider(minimum=0, maximum=1, step=0.05, label="Denoise | 去噪强度", value=1)
steps_c = gr.Slider(minimum=10, maximum=30, step=1, label="Sampling steps | 采样次数", value=20)
with gr.Row():
cfg_c = gr.Slider(minimum=0, maximum=20, step=0.1, label="CFG Scale | CFG权重", value=4)
seed_c = gr.Slider(minimum=-1, maximum=18446744073709550000, step=1, label="Seed | 种子数", value=-1)
with gr.Tab(label="Stage B 采样设置"):
with gr.Row():
sampler_name_b = gr.Dropdown(Choices.sampler, label="Sampling method | 采样方法", value=Choices.sampler[12])
scheduler_b = gr.Dropdown(Choices.scheduler, label="Schedule type | 采样计划表类型", value=Choices.scheduler[1])
with gr.Row():
denoise_b = gr.Slider(minimum=0, maximum=1, step=0.05, label="Denoise | 去噪强度", value=1)
steps_b = gr.Slider(minimum=4, maximum=12, step=1, label="Sampling steps | 采样次数", value=10)
with gr.Row():
cfg_b = gr.Slider(minimum=0, maximum=20, step=0.1, label="CFG Scale | CFG权重", value=1.1)
seed_b = gr.Slider(minimum=-1, maximum=18446744073709550000, step=1, label="Seed | 种子数", value=-1)
Postprocess.blocks("SC")
with gr.Column():
with gr.Row():
btn = gr.Button("Generate | 生成", elem_id="button")
btn2 = gr.Button("Interrupt | 终止")
output = gr.Gallery(preview=True, height=600)
with gr.Row():
SC.send_to_sd = gr.Button("发送图片至 SD")
SC.send_to_sc = gr.Button("发送图片至 SC")
if SVD.enable is True:
SC.send_to_svd = gr.Button("发送图片至 SVD")
SC.send_to_extras = gr.Button("发送图片至 Extras")
SC.send_to_info = gr.Button("发送图片至 Info")
SC.data = gr.State()
SC.index = gr.State()
btn.click(fn=SC.generate, inputs=[Initial.initialized, batch_count, positive_prompt, negative_prompt, width, height, batch_size, seed_c, steps_c, cfg_c, sampler_name_c, scheduler_c, denoise_c, seed_b, steps_b, cfg_b, sampler_name_b, scheduler_b, denoise_b, SC.input_image], outputs=[output, SC.data])
btn2.click(fn=Function.post_interrupt, inputs=None, outputs=None)
output.select(fn=Function.get_gallery_index, inputs=None, outputs=[SC.index])
class SVD:
if Default.design_mode == 1:
enable = True
elif "svd_xt_1_1.safetensors" in Choices.ckpt_list:
enable = True
else:
enable = False
def generate(input_image, width, height, video_frames, motion_bucket_id, fps, augmentation_level, min_cfg, seed, steps, cfg, sampler_name, scheduler, denoise, fps2, lossless, quality, method, progress=gr.Progress()):
ckpt_name = Function.get_model_path("svd_xt_1_1.safetensors")
seed = Function.gen_seed(seed)
if input_image is None:
return
else:
input_image = Function.upload_image(input_image)
workflow = {
"1": {"inputs": {"ckpt_name": ckpt_name}, "class_type": "ImageOnlyCheckpointLoader"},
"2": {"inputs": {"image": input_image, "upload": "image"}, "class_type": "LoadImage"},
"3": {"inputs": {"width": width, "height": height, "video_frames": video_frames, "motion_bucket_id": motion_bucket_id, "fps": fps, "augmentation_level": augmentation_level, "clip_vision": ["1", 1], "init_image": ["2", 0], "vae": ["1", 2]}, "class_type": "SVD_img2vid_Conditioning"},
"4": {"inputs": {"min_cfg": min_cfg, "model": ["1", 0]}, "class_type": "VideoLinearCFGGuidance"},
"5": {"inputs": {"seed": seed, "steps": steps, "cfg": cfg, "sampler_name": sampler_name, "scheduler": scheduler, "denoise": denoise, "model": ["4", 0], "positive": ["3", 0], "negative": ["3", 1], "latent_image": ["3", 2]}, "class_type": "KSampler"},
"6": {"inputs": {"samples": ["5", 0], "vae": ["1", 2]}, "class_type": "VAEDecode"},
"7": {"inputs": {"filename_prefix": "ComfyUI", "fps": fps2, "lossless": False, "quality": quality, "method": method, "images": ["6", 0]}, "class_type": "SaveAnimatedWEBP"}
}
return Function.gen_image(workflow, 1, 1, progress)[1]
def blocks():
with gr.Row():
with gr.Column():
SVD.input_image = gr.Image(value=None, type="pil")
with gr.Tab(label="基础设置"):
with gr.Row():
width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width | 图像宽度", value=512)
video_frames = gr.Slider(minimum=1, maximum=25, step=1, label="Video frames | 视频帧", value=25)
with gr.Row():
height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height | 图像高度", value=512)
fps = gr.Slider(minimum=1, maximum=30, step=1, label="FPS | 帧率", value=6)
with gr.Row():
with gr.Column():
augmentation_level = gr.Slider(minimum=0, maximum=1, step=0.01, label="Augmentation level | 增强级别", value=0)
motion_bucket_id = gr.Slider(minimum=1, maximum=256, step=1, label="Motion bucket id | 运动参数", value=127)
with gr.Column():
min_cfg = gr.Slider(minimum=0, maximum=20, step=0.5, label="Min CFG | 最小CFG权重", value=1)
with gr.Tab(label="采样设置"):
with gr.Row():
sampler_name = gr.Dropdown(Choices.sampler, label="Sampling method | 采样方法", value=Choices.sampler[12])
scheduler = gr.Dropdown(Choices.scheduler, label="Schedule type | 采样计划表类型", value=Choices.scheduler[1])
with gr.Row():
denoise = gr.Slider(minimum=0, maximum=1, step=0.05, label="Denoise | 去噪强度", value=1)
steps = gr.Slider(minimum=10, maximum=30, step=1, label="Sampling steps | 采样次数", value=20)
with gr.Row():
cfg = gr.Slider(minimum=0, maximum=20, step=0.1, label="CFG Scale | CFG权重", value=2.5)
seed = gr.Slider(minimum=-1, maximum=18446744073709550000, step=1, label="Seed | 种子数", value=-1)
with gr.Tab(label="输出设置"):
with gr.Row():
method = gr.Dropdown(["default", "fastest", "slowest"], label="Method | 输出方法", value="default")
lossless = gr.Dropdown(["true", "false"], label="Lossless | 无损压缩", value="false")
with gr.Row():
quality = gr.Slider(minimum=70, maximum=100, step=1, label="Quality | 输出质量", value=85)
fps2 = gr.Slider(minimum=1, maximum=30, step=1, label="FPS | 帧率", value=10)
with gr.Column():
with gr.Row():
btn = gr.Button("Generate | 生成", elem_id="button")
btn2 = gr.Button("Interrupt | 终止")
output = gr.Image(height=600)
btn.click(fn=SVD.generate, inputs=[SVD.input_image, width, height, video_frames, motion_bucket_id, fps, augmentation_level, min_cfg, seed, steps, cfg, sampler_name, scheduler, denoise, fps2, lossless, quality, method], outputs=[output])
btn2.click(fn=Function.post_interrupt, inputs=None, outputs=None)
class Extras:
def generate(initialized, input_image, progress=gr.Progress()):
module = "Extras"
if input_image is None:
return
else:
input_image = Function.upload_image(input_image)
workflow = {}
node_id = 1
workflow[str(node_id)] = {"inputs": {"image": input_image, "upload": "image"}, "class_type": "LoadImage"}
image_port = [str(node_id), 0]
if initialized is True:
if module not in Upscale.cache and module not in UpscaleWithModel.cache:
return
if module in Upscale.cache:
workflow, node_id, image_port = Upscale.add_node(module, workflow, node_id, image_port)
if module in UpscaleWithModel.cache:
workflow, node_id, image_port = UpscaleWithModel.add_node(module, workflow, node_id, image_port)
else:
return
node_id += 1
workflow[str(node_id)] = {"inputs": {"filename_prefix": "ComfyUI", "images": image_port}, "class_type": "SaveImage"}
output = Function.gen_image(workflow, 1, 1, progress)[0]
if output is not None:
output = output[0]
return output
def blocks():
with gr.Row():
with gr.Column():
Extras.input_image = gr.Image(value=None, type="pil")
with gr.Row():
with gr.Tab(label="算术放大"):
Upscale.blocks("Extras")
with gr.Row():
with gr.Tab(label="超分放大"):
UpscaleWithModel.blocks("Extras")
gr.HTML("注意：同时启用两种放大模式将先执行算术放大，再执行超分放大，最终放大倍数为二者放大倍数的乘积！")
with gr.Column():
with gr.Row():
btn = gr.Button("Generate | 生成", elem_id="button")
btn2 = gr.Button("Interrupt | 终止")
output = gr.Image(height=600)
btn.click(fn=Extras.generate, inputs=[Initial.initialized, Extras.input_image], outputs=[output])
btn2.click(fn=Function.post_interrupt, inputs=None, outputs=None)
class Info:
def generate(image_info, progress=gr.Progress()):
if not image_info or image_info is None or image_info == "仅支持API工作流！！！" or "Version:" in image_info or image_info == "None":
return
workflow = json.loads(image_info)
return Function.gen_image(workflow, 1, 1, progress)[0]
def order_workflow(workflow):
if workflow is None:
return gr.update(visible=False, value=None)
workflow = json.loads(workflow)
if "last_node_id" in workflow:
return gr.update(show_label=False, visible=True, value="仅支持API工作流！！！", lines=1)
workflow = Function.order_workflow(workflow)
lines = len(workflow) + 5
workflow_string = "{"
for node in workflow:
workflow_string = workflow_string + "\n" + f'"{node}": {workflow[node]},'
workflow_string = workflow_string + "\n}"
workflow_string = workflow_string.replace(",\n}", "\n}")
workflow_string = workflow_string.replace("'", '"')
return gr.update(label="Ordered workflow_api", show_label=True, visible=True, value=workflow_string, lines=lines)
def get_image_info(image_pil):
if image_pil is None:
return gr.update(visible=False, value=None)
else:
image_info = Function.get_image_info(image_pil)
if image_info == "None":
return gr.update(visible=False, value=None)
if "Version:" in image_info:
return gr.update(label="Image info", show_label=True, visible=True, value=image_info, lines=3)
return Info.order_workflow(image_info)
def hide_another_input(this_input):
if this_input is None:
return gr.update(visible=True)
return gr.update(visible=False)
def blocks():
with gr.Row():
with gr.Column():
Info.input_image = gr.Image(value=None, type="pil")
workflow = gr.File(label="workflow_api.json", file_types=[".json"], type="binary")
image_info = gr.Textbox(visible=False)
with gr.Column():
with gr.Row():
btn = gr.Button("Generate | 生成", elem_id="button")
btn2 = gr.Button("Interrupt | 终止")
output = gr.Gallery(preview=True, height=600)
btn.click(fn=Info.generate, inputs=[image_info], outputs=[output])
btn2.click(fn=Function.post_interrupt, inputs=None, outputs=None)
Info.input_image.change(fn=Info.hide_another_input, inputs=[Info.input_image], outputs=[workflow])
Info.input_image.change(fn=Info.get_image_info, inputs=[Info.input_image], outputs=[image_info])
workflow.change(fn=Info.hide_another_input, inputs=[workflow], outputs=[Info.input_image])
workflow.change(fn=Info.order_workflow, inputs=[workflow], outputs=[image_info])
with gr.Blocks(css="#button {background: #FFE1C0; color: #FF453A} .block.padded:not(.gradio-accordion) {padding: 0 !important;} div.form {border-width: 0; box-shadow: none; background: white; gap: 1.15em;}") as demo:
Initial.initialized = gr.Checkbox(value=False, visible=False)
with gr.Tab(label="Stable Diffusion"): SD.blocks()
if SC.enable is True:
with gr.Tab(label="Stable Cascade"): SC.blocks()
if SVD.enable is True:
with gr.Tab(label="Stable Video Diffusion"): SVD.blocks()
with gr.Tab(label="Extras"): Extras.blocks()
with gr.Tab(label="Info"): Info.blocks()
SD.send_to_sd.click(fn=Function.send_to, inputs=[SD.data, SD.index], outputs=[SD.input_image])
if SC.enable is True:
SD.send_to_sc.click(fn=Function.send_to, inputs=[SD.data, SD.index], outputs=[SC.input_image])
if SVD.enable is True:
SD.send_to_svd.click(fn=Function.send_to, inputs=[SD.data, SD.index], outputs=[SVD.input_image])
SD.send_to_extras.click(fn=Function.send_to, inputs=[SD.data, SD.index], outputs=[Extras.input_image])
SD.send_to_info.click(fn=Function.send_to, inputs=[SD.data, SD.index], outputs=[Info.input_image])
if SC.enable is True:
SC.send_to_sd.click(fn=Function.send_to, inputs=[SC.data, SC.index], outputs=[SD.input_image])
SC.send_to_sc.click(fn=Function.send_to, inputs=[SC.data, SC.index], outputs=[SC.input_image])
if SVD.enable is True:
SC.send_to_svd.click(fn=Function.send_to, inputs=[SC.data, SC.index], outputs=[SVD.input_image])
SC.send_to_extras.click(fn=Function.send_to, inputs=[SC.data, SC.index], outputs=[Extras.input_image])
SC.send_to_info.click(fn=Function.send_to, inputs=[SC.data, SC.index], outputs=[Info.input_image])
demo.queue(concurrency_count=100).launch(inbrowser=True, inline=False)