fluxhdupscaler / app.py
comrender's picture
Update app.py
33c9103 verified
raw
history blame
16.6 kB
import os
import random
import sys
from typing import Sequence, Mapping, Any, Union
import torch
import gradio as gr
from huggingface_hub import hf_hub_download
import spaces
# Download required models from Hugging Face
hf_hub_download(repo_id="black-forest-labs/FLUX.1-dev", filename="ae.safetensors", local_dir="models/vae")
hf_hub_download(repo_id="black-forest-labs/FLUX.1-dev", filename="flux1-dev.safetensors", local_dir="models/diffusion_models")
hf_hub_download(repo_id="comfyanonymous/flux_text_encoders", filename="clip_l.safetensors", local_dir="models/text_encoders")
hf_hub_download(repo_id="comfyanonymous/flux_text_encoders", filename="t5xxl_fp16.safetensors", local_dir="models/text_encoders")
hf_hub_download(repo_id="kim2091/UltraSharp", filename="4x-UltraSharp.pth", local_dir="models/upscale_models")
def get_value_at_index(obj: Union[Sequence, Mapping], index: int) -> Any:
"""Returns the value at the given index of a sequence or mapping."""
try:
return obj[index]
except KeyError:
return obj["result"][index]
def find_path(name: str, path: str = None) -> str:
"""Recursively looks at parent folders starting from the given path until it finds the given name."""
if path is None:
path = os.getcwd()
if name in os.listdir(path):
path_name = os.path.join(path, name)
print(f"{name} found: {path_name}")
return path_name
parent_directory = os.path.dirname(path)
if parent_directory == path:
return None
return find_path(name, parent_directory)
def add_comfyui_directory_to_sys_path() -> None:
"""Add 'ComfyUI' to the sys.path"""
comfyui_path = find_path("ComfyUI")
if comfyui_path is not None and os.path.isdir(comfyui_path):
sys.path.append(comfyui_path)
print(f"'{comfyui_path}' added to sys.path")
def add_extra_model_paths() -> None:
"""Parse the optional extra_model_paths.yaml file and add the parsed paths to the sys.path."""
try:
from main import load_extra_path_config
extra_model_paths = find_path("extra_model_paths.yaml")
if extra_model_paths is not None:
load_extra_path_config(extra_model_paths)
else:
print("Could not find the extra_model_paths config file.")
except ImportError:
try:
from utils.extra_config import load_extra_path_config
extra_model_paths = find_path("extra_model_paths.yaml")
if extra_model_paths is not None:
load_extra_path_config(extra_model_paths)
else:
print("Could not find the extra_model_paths config file.")
except ImportError:
print("Could not import extra config. Continuing without extra model paths.")
add_comfyui_directory_to_sys_path()
try:
add_extra_model_paths()
except Exception as e:
print(f"Warning: Could not load extra model paths: {e}")
def import_custom_nodes() -> None:
"""Find all custom nodes in the custom_nodes folder and add those node objects to NODE_CLASS_MAPPINGS"""
try:
import asyncio
import execution
from nodes import init_extra_nodes
import server
# Check if we're already in an event loop
try:
loop = asyncio.get_event_loop()
if loop.is_running():
# We're in an existing loop, use it
pass
else:
# Loop exists but not running, set a new one
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
except RuntimeError:
# No loop exists, create one
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
server_instance = server.PromptServer(loop)
execution.PromptQueue(server_instance)
init_extra_nodes()
except Exception as e:
print(f"Warning: Could not initialize custom nodes: {e}")
print("Continuing with basic ComfyUI nodes only...")
from nodes import NODE_CLASS_MAPPINGS
# Pre-load models outside the decorated function for ZeroGPU efficiency
try:
import_custom_nodes()
# Initialize model loaders
dualcliploader = NODE_CLASS_MAPPINGS["DualCLIPLoader"]()
dualcliploader_54 = dualcliploader.load_clip(
clip_name1="clip_l.safetensors",
clip_name2="t5xxl_fp16.safetensors",
type="flux",
device="default",
)
upscalemodelloader = NODE_CLASS_MAPPINGS["UpscaleModelLoader"]()
upscalemodelloader_44 = upscalemodelloader.load_model(model_name="4x-UltraSharp.pth")
vaeloader = NODE_CLASS_MAPPINGS["VAELoader"]()
vaeloader_55 = vaeloader.load_vae(vae_name="ae.safetensors")
unetloader = NODE_CLASS_MAPPINGS["UNETLoader"]()
unetloader_58 = unetloader.load_unet(
unet_name="flux1-dev.safetensors", weight_dtype="default"
)
downloadandloadflorence2model = NODE_CLASS_MAPPINGS["DownloadAndLoadFlorence2Model"]()
downloadandloadflorence2model_52 = downloadandloadflorence2model.loadmodel(
model="microsoft/Florence-2-large", precision="fp16", attention="sdpa"
)
# Pre-load models to GPU for efficiency
try:
from comfy import model_management
model_loaders = [dualcliploader_54, vaeloader_55, unetloader_58, downloadandloadflorence2model_52]
valid_models = [
getattr(loader[0], 'patcher', loader[0])
for loader in model_loaders
if not isinstance(loader[0], dict) and not isinstance(getattr(loader[0], 'patcher', None), dict)
]
model_management.load_models_gpu(valid_models)
print("Models successfully pre-loaded to GPU")
except Exception as e:
print(f"Warning: Could not pre-load models to GPU: {e}")
print("ComfyUI setup completed successfully!")
except Exception as e:
print(f"Error during ComfyUI setup: {e}")
print("Please check that all required custom nodes are installed.")
raise
@spaces.GPU(duration=120) # Adjust duration based on your workflow speed
def enhance_image(image_input, upscale_factor, steps, cfg_scale, denoise_strength, guidance_scale):
"""
Main function to enhance and upscale images using Florence-2 captioning and FLUX upscaling
"""
try:
with torch.inference_mode():
# Handle different input types (file upload vs URL)
if isinstance(image_input, str) and image_input.startswith(('http://', 'https://')):
# Load from URL
load_image_from_url_mtb = NODE_CLASS_MAPPINGS["Load Image From Url (mtb)"]()
load_image_result = load_image_from_url_mtb.load(url=image_input)
else:
# Load from uploaded file
loadimage = NODE_CLASS_MAPPINGS["LoadImage"]()
load_image_result = loadimage.load_image(image=image_input)
# Generate detailed caption using Florence-2
florence2run = NODE_CLASS_MAPPINGS["Florence2Run"]()
florence2run_51 = florence2run.encode(
text_input="",
task="more_detailed_caption",
fill_mask=True,
keep_model_loaded=False,
max_new_tokens=1024,
num_beams=3,
do_sample=True,
output_mask_select="",
seed=random.randint(1, 2**64),
image=get_value_at_index(load_image_result, 0),
florence2_model=get_value_at_index(downloadandloadflorence2model_52, 0),
)
# Encode the generated caption
cliptextencode = NODE_CLASS_MAPPINGS["CLIPTextEncode"]()
cliptextencode_6 = cliptextencode.encode(
text=get_value_at_index(florence2run_51, 2),
clip=get_value_at_index(dualcliploader_54, 0),
)
# Encode empty negative prompt
cliptextencode_42 = cliptextencode.encode(
text="", clip=get_value_at_index(dualcliploader_54, 0)
)
# Set up upscale factor
primitivefloat = NODE_CLASS_MAPPINGS["PrimitiveFloat"]()
primitivefloat_60 = primitivefloat.execute(value=upscale_factor)
# Apply FLUX guidance
fluxguidance = NODE_CLASS_MAPPINGS["FluxGuidance"]()
fluxguidance_26 = fluxguidance.append(
guidance=guidance_scale,
conditioning=get_value_at_index(cliptextencode_6, 0)
)
# Perform ultimate upscaling
ultimatesdupscale = NODE_CLASS_MAPPINGS["UltimateSDUpscale"]()
ultimatesdupscale_50 = ultimatesdupscale.upscale(
upscale_by=get_value_at_index(primitivefloat_60, 0),
seed=random.randint(1, 2**64),
steps=steps,
cfg=cfg_scale,
sampler_name="euler",
scheduler="normal",
denoise=denoise_strength,
mode_type="Linear",
tile_width=1024,
tile_height=1024,
mask_blur=8,
tile_padding=32,
seam_fix_mode="None",
seam_fix_denoise=1,
seam_fix_width=64,
seam_fix_mask_blur=8,
seam_fix_padding=16,
force_uniform_tiles=True,
tiled_decode=False,
image=get_value_at_index(load_image_result, 0),
model=get_value_at_index(unetloader_58, 0),
positive=get_value_at_index(fluxguidance_26, 0),
negative=get_value_at_index(cliptextencode_42, 0),
vae=get_value_at_index(vaeloader_55, 0),
upscale_model=get_value_at_index(upscalemodelloader_44, 0),
)
# Save the result
saveimage = NODE_CLASS_MAPPINGS["SaveImage"]()
saveimage_43 = saveimage.save_images(
filename_prefix="enhanced_image",
images=get_value_at_index(ultimatesdupscale_50, 0),
)
# Return the path to the saved image
saved_path = f"output/{saveimage_43['ui']['images'][0]['filename']}"
# Also return the generated caption for user feedback
generated_caption = get_value_at_index(florence2run_51, 2)
return saved_path, generated_caption
except Exception as e:
print(f"Error in enhance_image: {str(e)}")
raise gr.Error(f"Enhancement failed: {str(e)}")
# Create the Gradio interface
def create_interface():
with gr.Blocks(
title="πŸš€ AI Image Enhancer - Florence-2 + FLUX",
theme=gr.themes.Soft(),
css="""
.gradio-container {
max-width: 1200px !important;
}
.main-header {
text-align: center;
margin-bottom: 2rem;
}
.result-gallery {
min-height: 400px;
}
"""
) as app:
gr.HTML("""
<div class="main-header">
<h1>🎨 AI Image Enhancer</h1>
<p>Upload an image or provide a URL to enhance it using Florence-2 captioning and FLUX upscaling</p>
</div>
""")
with gr.Row():
with gr.Column(scale=1):
gr.HTML("<h3>πŸ“€ Input Settings</h3>")
with gr.Tabs():
with gr.TabItem("πŸ“ Upload Image"):
image_upload = gr.Image(
label="Upload Image",
type="filepath",
height=300
)
with gr.TabItem("πŸ”— Image URL"):
image_url = gr.Textbox(
label="Image URL",
placeholder="https://example.com/image.jpg",
value="https://upload.wikimedia.org/wikipedia/commons/thumb/a/a7/Example.jpg/800px-Example.jpg"
)
gr.HTML("<h3>βš™οΈ Enhancement Settings</h3>")
upscale_factor = gr.Slider(
minimum=1.0,
maximum=4.0,
value=2.0,
step=0.5,
label="Upscale Factor",
info="How much to upscale the image"
)
steps = gr.Slider(
minimum=10,
maximum=50,
value=25,
step=5,
label="Steps",
info="Number of denoising steps"
)
cfg_scale = gr.Slider(
minimum=0.5,
maximum=10.0,
value=1.0,
step=0.5,
label="CFG Scale",
info="Classifier-free guidance scale"
)
denoise_strength = gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.3,
step=0.1,
label="Denoise Strength",
info="How much to denoise the image"
)
guidance_scale = gr.Slider(
minimum=1.0,
maximum=10.0,
value=3.5,
step=0.5,
label="Guidance Scale",
info="FLUX guidance strength"
)
enhance_btn = gr.Button(
"πŸš€ Enhance Image",
variant="primary",
size="lg"
)
with gr.Column(scale=1):
gr.HTML("<h3>πŸ“Š Results</h3>")
output_image = gr.Image(
label="Enhanced Image",
type="filepath",
height=400,
interactive=False
)
generated_caption = gr.Textbox(
label="Generated Caption",
placeholder="The AI-generated caption will appear here...",
lines=3,
interactive=False
)
gr.HTML("""
<div style="margin-top: 1rem; padding: 1rem; background: #f0f0f0; border-radius: 8px;">
<h4>πŸ’‘ How it works:</h4>
<ol>
<li>Florence-2 analyzes your image and generates a detailed caption</li>
<li>FLUX uses this caption to guide the upscaling process</li>
<li>The result is an enhanced, higher-resolution image</li>
</ol>
</div>
""")
# Event handlers
def process_image(img_upload, img_url, upscale_f, steps_val, cfg_val, denoise_val, guidance_val):
# Determine input source
image_input = img_upload if img_upload is not None else img_url
if not image_input:
raise gr.Error("Please provide an image (upload or URL)")
return enhance_image(image_input, upscale_f, steps_val, cfg_val, denoise_val, guidance_val)
enhance_btn.click(
fn=process_image,
inputs=[
image_upload,
image_url,
upscale_factor,
steps,
cfg_scale,
denoise_strength,
guidance_scale
],
outputs=[output_image, generated_caption]
)
# Example inputs
gr.Examples(
examples=[
[None, "https://upload.wikimedia.org/wikipedia/commons/thumb/a/a7/Example.jpg/800px-Example.jpg", 2.0, 25, 1.0, 0.3, 3.5],
[None, "https://picsum.photos/512/512", 2.0, 20, 1.5, 0.4, 4.0],
],
inputs=[
image_upload,
image_url,
upscale_factor,
steps,
cfg_scale,
denoise_strength,
guidance_scale
]
)
return app
if __name__ == "__main__":
app = create_interface()
app.launch(share=True, server_name="0.0.0.0", server_port=7860)