app.py

import gradio as gr
import roop.globals
from roop.core import (
    start,
    decode_execution_providers,
    suggest_max_memory,
    suggest_execution_threads,
)
from roop.processors.frame.core import get_frame_processors_modules
from roop.utilities import normalize_output_path
from PIL import Image
import io

def process_face_swap(source_image, target_image, apply_face_enhancer):
    # Convert images to PIL Image objects and save them to in-memory file-like objects
    src_image_bytes = io.BytesIO()
    tgt_image_bytes = io.BytesIO()
    
    source_image.save(src_image_bytes, format='JPEG')
    target_image.save(tgt_image_bytes, format='JPEG')

    src_image_bytes.seek(0)
    tgt_image_bytes.seek(0)

    src_image_path = "source_image.jpg"
    tgt_image_path = "target_image.jpg"
    output_path = "result_image.jpg"

    with open(src_image_path, 'wb') as f:
        f.write(src_image_bytes.getvalue())
    
    with open(tgt_image_path, 'wb') as f:
        f.write(tgt_image_bytes.getvalue())

    # Set Roop configuration
    roop.globals.source_path = src_image_path
    roop.globals.target_path = tgt_image_path
    roop.globals.output_path = normalize_output_path(
        roop.globals.source_path, roop.globals.target_path, output_path
    )
    roop.globals.frame_processors = ["face_swapper", "face_enhancer"] if apply_face_enhancer else ["face_swapper"]
    roop.globals.headless = True
    roop.globals.keep_fps = True
    roop.globals.keep_audio = True
    roop.globals.keep_frames = False
    roop.globals.many_faces = False
    roop.globals.video_encoder = "libx264"
    roop.globals.video_quality = 18
    roop.globals.max_memory = suggest_max_memory()
    roop.globals.execution_providers = decode_execution_providers(["cuda", "cpu"])
    roop.globals.execution_threads = suggest_execution_threads()

    # Print paths for debugging
    print(f"Starting the process with:")
    print(f"Source: {roop.globals.source_path}")
    print(f"Target: {roop.globals.target_path}")
    print(f"Output: {roop.globals.output_path}")

    # Execute face swap process
    for processor in get_frame_processors_modules(roop.globals.frame_processors):
        if not processor.pre_check():
            return

    start()
    
    # Load and return the output image
    with open(roop.globals.output_path, 'rb') as f:
        output_image = f.read()
    
    return output_image

# Create Gradio interface
iface = gr.Interface(
    fn=process_face_swap,
    inputs=[
        gr.Image(type="pil", label="Source Image"),
        gr.Image(type="pil", label="Target Image"),
        gr.Checkbox(label="Apply Face Enhancer")
    ],
    outputs=gr.Image(type="pil"),
    description="Swap faces between two images. Use the face enhancer option to improve the result. Processing may take longer if you use Face Enhancer. Becose this Modle is running using CPU."
)

iface.launch()