README.md

metadata

title: Multi-Mixture Speaker Identification
emoji: 🗣️
colorFrom: gray
colorTo: blue
sdk: gradio
sdk_version: 6.3.0
python_version: '3.11'
app_file: app.py
pinned: false
models:
  - HiMind/Multi-Mixture_Speaker_ID

MMM Speaker Identification

Interactive demo of a research-oriented hybrid model (VAE + RNN + HMM + GMM + Transformer) for speaker identification from short audio clips. Upload speaker examples to enroll them and upload a query clip to identify the most likely speaker based on learned latent embeddings.

Designed and trained by Chance Brownfield.

How to use

1. Under Speaker A / B / C, upload short mono audio files (sample rate should match the model setup; 1–5 seconds is typical).
2. Upload a Query Audio file.
3. Click Identify Speaker. The app returns the predicted speaker label and per-speaker scores (log-likelihoods).

Author

Chance Brownfield
Designer and trainer of the MMM architecture
Email: HiMindAi@proton.me