This project is still working but some stuff is deprecated

[CyberT33N]

First of all thank you for this amazing project. Secondly everything is still working but some stuff is deprecated. Here is how I make everything running:

• https://github.com/CyberT33N/riffusion-cheat-sheet/blob/main/README.md

Duplicated of:

• https://github.com/riffusion/riffusion-hobby/pull/165

However, maybe still usefully for people who want to use the UI

---

Install

• https://github.com/riffusion/riffusion-hobby
• https://github.com/riffusion/riffusion-app-hobby

Clone riffusion-hobby & riffusion-app-hobby

cd ~/Projects/ai/audio

git clone https://github.com/riffusion/riffusion-hobby.git
git clone https://github.com/riffusion/riffusion-app-hobby.git

Because the head commit is deprecated of riffusion-hobby and the latest torchaudio of requirements.txt is not supported you must modify self.inverse_mel_scaler:

• ~/Projects/ai/audio/riffusion-hobby/riffusion/spectrogram_converter.py

# https://pytorch.org/audio/stable/generated/torchaudio.transforms.InverseMelScale.html
self.inverse_mel_scaler = torchaudio.transforms.InverseMelScale(
    n_stft=params.n_fft // 2 + 1,
    n_mels=params.num_frequencies,
    sample_rate=params.sample_rate,
    f_min=params.min_frequency,
    f_max=params.max_frequency,
    norm=params.mel_scale_norm,
    mel_scale=params.mel_scale_type,
).to(self.device)

Install and start riffusion-hobby:

cd ~/Projects/ai/audio/riffusion-hobby

# pyenv install 3.9
pyenv local 3.9

python3 -m venv venv
source venv/bin/activate

sudo apt install sox

# Those dependencies are missing in the head commit of requirements.txt
pip install numpy typing_extensions

python -m pip install -r requirements.txt

# cache_download not longer supported in latest version so we downgrade
pip install huggingface_hub==0.24.7

Install and start riffusion-app-hobby. Create .env.local and insert RIFFUSION_FLASK_URL=http://127.0.0.1:3013/run_inference/. Then run:

cd ~/Projects/ai/audio/riffusion-app-hobby

npm i
npm run dev

[amuhareb]

Thanks for the write up, I look forward to trying it out