---
license: mpl-2.0
datasets:
- CommonVoice
- LibriSpeech
- LibriTTS
- VCTK
- VoxCeleb
tags:
- speaker-encoding
- speaker-embedding
- speaker-verification
- speaker
- speech
library_name: PyTorch
---

This repository contains released Mozilla SpeakerEncoder models.

Due to original codebase migration and SpeakerEncoder depracation, I decided to extract inference code into my own repository. For ease of integration, I also release relevant models:
- `model_mozilla_128.pth.tar` - which is [Speaker-Encoder-iter25k](https://drive.google.com/open?id=1CwXxwm0qKk2lcPL_4EQDirD8XMyvHIqR) that was trained on LibriSpeech,
- `model_mozilla_256.pth.tar` - which is [Speaker-Encoder](https://drive.google.com/drive/folders/15oeBYf6Qn1edONkVLXe82MzdIi3O_9m3?usp=sharing) by [@mueller91](https://github.com/mueller91) that was trained on LibriTTS, VCTK, VoxCeleb and CommonVoice.

Original SpeakerEncoder implementation is available [here](https://github.com/coqui-ai/TTS/tree/f9001a4bdd8a169da8b4eefd481461f370def8f9/TTS/speaker_encoder) and originally released models are available [here](https://github.com/mozilla/TTS/wiki/Released-Models). If you are interested in the current version of Coqui.ai TTS go [here](https://github.com/coqui-ai/TTS).