INNER CODE UNIT · Python

inference

jdh-algo/JoyHallo · scripts/app.py:312

def inference(cfg, image_processor, audio_processor, pipeline, audioproj, save_dir, global_step, clip_length):
    ref_img_path = cfg.ref_img_path
    audio_path = cfg.audio_path
    source_image_pixels, \
    source_image_face_region, \
    source_image_face_emb, \
    source_image_full_mask, \
    source_image_face_mask, \
    source_image_lip_mask = image_processor.preprocess(
        ref_img_path, os.path.join(save_dir, '.cache'), cfg.face_expand_ratio)
    audio_emb, audio_length = audio_processor.preprocess(
        audio_path, clip_length)

    audio_emb = process_audio_emb(audio_emb)

    source_image_pixels = source_image_pixels.unsqueeze(0)
    source_image_face_region = source_image_face_region.unsqueeze(0)
    source_image_face_emb = source_image_face_emb.reshape(1, -1)

View source record →