Instructions to use DevParker/VibeVoice7b-low-vram with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- VibeVoice
How to use DevParker/VibeVoice7b-low-vram with VibeVoice:
import torch, soundfile as sf, librosa, numpy as np from vibevoice.processor.vibevoice_processor import VibeVoiceProcessor from vibevoice.modular.modeling_vibevoice_inference import VibeVoiceForConditionalGenerationInference # Load voice sample (should be 24kHz mono) voice, sr = sf.read("path/to/voice_sample.wav") if voice.ndim > 1: voice = voice.mean(axis=1) if sr != 24000: voice = librosa.resample(voice, sr, 24000) processor = VibeVoiceProcessor.from_pretrained("DevParker/VibeVoice7b-low-vram") model = VibeVoiceForConditionalGenerationInference.from_pretrained( "DevParker/VibeVoice7b-low-vram", torch_dtype=torch.bfloat16 ).to("cuda").eval() model.set_ddpm_inference_steps(5) inputs = processor(text=["Speaker 0: Hello!\nSpeaker 1: Hi there!"], voice_samples=[[voice]], return_tensors="pt") audio = model.generate(**inputs, cfg_scale=1.3, tokenizer=processor.tokenizer).speech_outputs[0] sf.write("output.wav", audio.cpu().numpy().squeeze(), 24000) - Notebooks
- Google Colab
- Kaggle
When is the new fp8 version coming?
#4
by ca231321 - opened
When is the new fp8 version coming? Curious to see how good it is.
+1 thanks for the 4 bit is awesome, the 8 bit will be sublime.
Super anxious to try this. I know there are other projects that have already integrated the 4bit. And I have tested it a good bit. While good.... I think the 8bit will work even better with greater accuracy. I have a 12 GB card. Do you need a tester by chance? I would be willing to help test it if you need.
I am getting the feeling this 8bit version is not coming any more. It's been two weeks now and nothing.
ditto~ but let's give the gentleman some more time...
please don't include 8 bit in the model card if it's not yet available