(audio_path, checkpoint, device)
| 37 | |
| 38 | |
| 39 | def get_audio_emb(audio_path, checkpoint, device): |
| 40 | audio, audio_rate = torchaudio.load(audio_path, channels_first=False) |
| 41 | assert audio_rate == 16000, 'Only 16 kHZ audio is supported.' |
| 42 | audio = audio[None, None, :, 0].to(device) |
| 43 | |
| 44 | audio_encoder = AudioEncoder(checkpoint).to(device) |
| 45 | |
| 46 | emb = audio_encoder(audio) |
| 47 | return audio, emb |
| 48 | |
| 49 | |
| 50 | def get_id_frame(path, random=False, resize=128): |