kicks-api / README.md
thepaul's picture
Claude Opus 4.8
Docs: correct vocoder description and document /health endpoint
c9c8331
|
Raw History Blame Contribute Delete
677 Bytes
---
title: Kicks API
emoji: 🥁
colorFrom: purple
colorTo: blue
sdk: docker
app_port: 7860
---
# Kicks API
VAE-powered kick drum synthesizer API (FastAPI). Decodes five PCA sliders through
the VAE's latent space into a spectrogram, then to 44.1 kHz audio. Uses the
BigVGAN v2 vocoder on GPU, with a Griffin-LIM fallback on CPU.
## API Endpoints
- `GET /config` — Returns slider definitions (PC names, ranges)
- `GET /generate?pc1=0.5&pc2=0.5&pc3=0.5&pc4=0.5&pc5=0.5` — Returns WAV audio
- `GET /spectrogram?pc1=0.5&pc2=0.5&pc3=0.5&pc4=0.5&pc5=0.5` — Returns spectrogram JSON
- `GET /health` — Liveness/readiness check
Slider values are 0.0–1.0. Default is 0.5.