cyan knoll
Hey everyone! I’m excited to share SESA Fast Separation, an alternative to SESA Audio Separation. This audio separation tool splits music into vocals and instrumentals, offering a simple, high-quality experience. I built it with AI help despite no coding skills, and I’d love your feedback! 😊
(This Colab runs much faster compared to SESA Audio Separation.)
What’s SESA Fast Separation?
Designed for music producers and audio enthusiasts, this alternative to SESA Audio Separation lets you experiment with sound creatively.
- Audio Separation: Split vocals and instrumentals with Roformer models:
- MelBand Roformer Kim | FT 3 by unwa
- MelBand Roformer | Karaoke by becruily
- MelBand Roformer | FVX by Gabox
- MelBand Roformer | INSTV8N by Gabox
- MelBand Roformer | INSTV8 by Gabox
- MelBand Roformer | INSTV7N by Gabox
- MelBand Roformer | Instrumental Bleedless V3 by Gabox
- MelBand Roformer Kim | Inst V1 (E) Plus by Unwa
- MelBand Roformer Kim | Inst V1 Plus by Unwa
- All Roformer models are supported!
- Coming Soon: Apollo Audio Enhancement: For cleaner sound.
- Config Saving: Save settings feature in progress.
- Auto Ensemble: Combine models using average, median, max, or min methods.
- The ability to specify stems you don't want.
- User-Friendly Interface: Gradio with progress bars and logs.
- Output Formats: WAV, FLAC, MP3, OGG, and more.
- Colab-Based: Easy access on Google Colab.
Tech Used (With AI Help)
I don’t code, so AI assisted me:
- Languages & Libraries: Python, Gradio, Librosa, Soundfile, PyTorch, yt-dlp.
- Tools: UVR-based models, ensemble processing.
- Features: tqdm for progress, segment size, overlap, pitch shift.
Try It Out!
Check out SESA Fast Separation here:
[https://colab.research.google.com/drive/1u9oUj3T0Z5F-Jl3J6Nm0dtAGsEESx80F?usp=sharing]
