Audiocraft is a library for audio processing and generation with deep learning. It features the state-of-the-art EnCodec audio compressor / tokenizer, along with MusicGen, a simple and controllable music generation LM with textual and melodic conditioning.
MIT License
20.18k
stars
2.01k
forks
source link
Generating audio samples from the same seed ? #373
Hello I am using the following workflow to generate audio samples using audio craft. Basically I am outputting from the llm (in this case Zephyr 7b mistral) to be used as an input. I am wondering how can i
Set the seed so the generations are consistent.
2 Is there a way to generate longer samples for example having the same seed
Generate an intro for 30 s with xyz
Generate verse for 30s with zlm
The workflow is listed below
!pip install -q --upgrade huggingface_hub git+https://github.com/huggingface/transformers.git
client = InferenceClient(model="HuggingFaceH4/zephyr-7b-beta",
token=HF_TOKEN)
prompt = "Dark industrial piano set in the 1960s"
additional_prompt = "Imagine you are a pianist."
new_prompt = additional_prompt + " " + prompt
input = f"Take the next sentence and enrich it feeling, keep it compact. {prompt}"
output = client.text_generation(input, max_new_tokens = 50)
print(output)
import torch
from transformers import pipeline
vibes = pipeline("text-to-audio",
"facebook/musicgen-stereo-medium",
torch_dtype=torch.float16,
device="cuda")
music_pipe = pipeline("text-to-audio", model="facebook/musicgen-small")
out = vibes(oufrom IPython.display import Audio
Audio(out["audio"][0], rate=32000))
Hello I am using the following workflow to generate audio samples using audio craft. Basically I am outputting from the llm (in this case Zephyr 7b mistral) to be used as an input. I am wondering how can i
The workflow is listed below
Any tips would be highly appreciated