worker_runpod_gradio.py

import os, random, time
import torch
from diffusers import AutoencoderKL, FlowMatchEulerDiscreteScheduler
from tqdm import tqdm

from memo.models.audio_proj import AudioProjModel
from memo.models.image_proj import ImageProjModel
from memo.models.unet_2d_condition import UNet2DConditionModel
from memo.models.unet_3d import UNet3DConditionModel
from memo.pipelines.video_pipeline import VideoPipeline
from memo.utils.audio_utils import extract_audio_emotion_labels, preprocess_audio, resample_audio
from memo.utils.vision_utils import preprocess_image, tensor_to_video

device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
weight_dtype = torch.bfloat16

with torch.inference_mode():
    vae = AutoencoderKL.from_pretrained("/content/memo/checkpoints/vae").to(device=device, dtype=weight_dtype)
    reference_net = UNet2DConditionModel.from_pretrained("/content/memo/checkpoints", subfolder="reference_net", use_safetensors=True)
    diffusion_net = UNet3DConditionModel.from_pretrained("/content/memo/checkpoints", subfolder="diffusion_net", use_safetensors=True)
    image_proj = ImageProjModel.from_pretrained("/content/memo/checkpoints", subfolder="image_proj", use_safetensors=True)
    audio_proj = AudioProjModel.from_pretrained("/content/memo/checkpoints", subfolder="audio_proj", use_safetensors=True)

    vae.requires_grad_(False).eval()
    reference_net.requires_grad_(False).eval()
    diffusion_net.requires_grad_(False).eval()
    image_proj.requires_grad_(False).eval()
    audio_proj.requires_grad_(False).eval()
    reference_net.enable_xformers_memory_efficient_attention()
    diffusion_net.enable_xformers_memory_efficient_attention()

    noise_scheduler = FlowMatchEulerDiscreteScheduler()
    pipeline = VideoPipeline(vae=vae, reference_net=reference_net, diffusion_net=diffusion_net, scheduler=noise_scheduler, image_proj=image_proj)
    pipeline.to(device=device, dtype=weight_dtype)

@torch.inference_mode()
def generate(input_video, input_audio, seed):
    resolution = 512
    num_generated_frames_per_clip = 16
    fps = 30
    num_init_past_frames = 2
    num_past_frames = 16
    inference_steps = 20
    cfg_scale = 3.5

    if seed == 0:
        random.seed(int(time.time()))
        seed = random.randint(0, 18446744073709551615)

    generator = torch.manual_seed(seed)
    img_size = (resolution, resolution)
    pixel_values, face_emb = preprocess_image(face_analysis_model="/content/memo/checkpoints/misc/face_analysis", image_path=input_video, image_size=resolution)

    output_dir = "/content/memo/outputs"
    os.makedirs(output_dir, exist_ok=True)
    cache_dir = os.path.join(output_dir, "audio_preprocess")
    os.makedirs(cache_dir, exist_ok=True)
    input_audio = resample_audio(input_audio, os.path.join(cache_dir, f"{os.path.basename(input_audio).split('.')[0]}-16k.wav"))

    audio_emb, audio_length = preprocess_audio(
        wav_path=input_audio,
        num_generated_frames_per_clip=num_generated_frames_per_clip,
        fps=fps,
        wav2vec_model="/content/memo/checkpoints/wav2vec2",
        vocal_separator_model="/content/memo/checkpoints/misc/vocal_separator/Kim_Vocal_2.onnx",
        cache_dir=cache_dir,
        device=device,
    )
    audio_emotion, num_emotion_classes = extract_audio_emotion_labels(
        model="/content/memo/checkpoints",
        wav_path=input_audio,
        emotion2vec_model="/content/memo/checkpoints/emotion2vec_plus_large",
        audio_length=audio_length,
        device=device,
    )

    video_frames = []
    num_clips = audio_emb.shape[0] // num_generated_frames_per_clip
    for t in tqdm(range(num_clips), desc="Generating video clips"):
        if len(video_frames) == 0:
            past_frames = pixel_values.repeat(num_init_past_frames, 1, 1, 1)
            past_frames = past_frames.to(dtype=pixel_values.dtype, device=pixel_values.device)
            pixel_values_ref_img = torch.cat([pixel_values, past_frames], dim=0)
        else:
            past_frames = video_frames[-1][0]
            past_frames = past_frames.permute(1, 0, 2, 3)
            past_frames = past_frames[0 - num_past_frames :]
            past_frames = past_frames * 2.0 - 1.0
            past_frames = past_frames.to(dtype=pixel_values.dtype, device=pixel_values.device)
            pixel_values_ref_img = torch.cat([pixel_values, past_frames], dim=0)

        pixel_values_ref_img = pixel_values_ref_img.unsqueeze(0)
        audio_tensor = (audio_emb[t * num_generated_frames_per_clip : min((t + 1) * num_generated_frames_per_clip, audio_emb.shape[0])].unsqueeze(0).to(device=audio_proj.device, dtype=audio_proj.dtype))
        audio_tensor = audio_proj(audio_tensor)
        audio_emotion_tensor = audio_emotion[t * num_generated_frames_per_clip : min((t + 1) * num_generated_frames_per_clip, audio_emb.shape[0])]

        pipeline_output = pipeline(
            ref_image=pixel_values_ref_img,
            audio_tensor=audio_tensor,
            audio_emotion=audio_emotion_tensor,
            emotion_class_num=num_emotion_classes,
            face_emb=face_emb,
            width=img_size[0],
            height=img_size[1],
            video_length=num_generated_frames_per_clip,
            num_inference_steps=inference_steps,
            guidance_scale=cfg_scale,
            generator=generator,
        )
        video_frames.append(pipeline_output.videos)

    video_frames = torch.cat(video_frames, dim=2)
    video_frames = video_frames.squeeze(0)
    video_frames = video_frames[:, :audio_length]

    video_path = f"/content/memo-{seed}-tost.mp4"
    tensor_to_video(video_frames, video_path, input_audio, fps=fps)

    return video_path

import gradio as gr

with gr.Blocks(css=".gradio-container {max-width: 1080px !important}", analytics_enabled=False) as demo:
    with gr.Row():
        with gr.Column():
            input_video = gr.Image(label="Upload Input Image", type="filepath")
            input_audio = gr.Audio(label="Upload Input Audio", type="filepath")
            seed = gr.Number(label="Seed (0 for Random)", value=0, precision=0)
        with gr.Column():
            video_output = gr.Video(label="Generated Video")
            generate_button = gr.Button("Generate")

    generate_button.click(
        fn=generate,
        inputs=[input_video, input_audio, seed],
        outputs=[video_output],
    )

demo.queue().launch(inline=False, share=False, debug=True, server_name='0.0.0.0', server_port=7860, allowed_paths=["/content"])