Spaces:
Running
Running
File size: 1,496 Bytes
c17b696 d533c9c c17b696 8c2e759 d533c9c 600e950 2b5d8ed 8c2e759 d533c9c c17b696 d533c9c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 |
from io import BytesIO
import streamlit as st
import soundfile as sf
from librosa.util import normalize
from librosa.beat import beat_track
from audiodiffusion import AudioDiffusion
if __name__ == "__main__":
st.header("Audio Diffusion")
st.markdown(
"Generate audio using Huggingface diffusers.\
The models without 'latent' or 'ddim' give better results but take about \
20 minutes without a GPU.", )
model_id = st.selectbox("Model", [
"teticio/audio-diffusion-256", "teticio/audio-diffusion-breaks-256",
"teticio/audio-diffusion-instrumental-hiphop-256",
"teticio/audio-diffusion-ddim-256",
"teticio/latent-audio-diffusion-256",
"teticio/latent-audio-diffusion-ddim-256"
],
index=5)
audio_diffusion = AudioDiffusion(model_id=model_id)
if st.button("Generate"):
st.markdown("Generating...")
image, (sample_rate,
audio) = audio_diffusion.generate_spectrogram_and_audio()
st.image(image, caption="Mel spectrogram")
buffer = BytesIO()
sf.write(buffer, normalize(audio), sample_rate, format="WAV")
st.audio(buffer, format="audio/wav")
audio = AudioDiffusion.loop_it(audio, sample_rate)
if audio is not None:
st.markdown("Loop")
buffer = BytesIO()
sf.write(buffer, normalize(audio), sample_rate, format="WAV")
st.audio(buffer, format="audio/wav")
|