A PyTorch-based Speech Toolkit

The SpeechBrain project aims to build a novel speech toolkit fully based on PyTorch. With SpeechBrain users can easily create speech processing systems, ranging from speech recognition (both HMM/DNN and end-to-end), speaker recognition, speech enhancement, speech separation, multi-microphone speech processing, and many others.

This repository contains the SpeechBrain Benchmarks

Transcription and TTS Rest API (OpenAI Whisper, Speechbrain)

Extensions to YAML syntax for better python interaction

基于speechbrain的快速目标说话人提取TSE服务

[ISMIR 2022] Transfer Learning of wav2vec 2.0 for Automatic Lyric Transcription

[TOMM 2024] Automatic Lyric Transcription and Automatic Music Transcription from Multimodal Singing

Real-time speaker diarization using straightforward, intuitive logic - High accuracy thanks to SpeechBrain/Pyannote-WeSpeaker models

StutterFormer is an AI model that aims to be able to receive a speech sample with stuttering disfluencies, and return it with the disfluencies attenuated or eliminated.

Text-to-Speech Recipe Users can create speech signals from an input text by using text-to-speech (TTS), also referred to as speech synthesis. Popular TTS and Vocoder models, such as Tacotron 2, are supported by SpeechBrain (e.g, HiFIGAN).

Backend of anti-fraud system based on speaker identification technology. 基于声纹识别的反诈系统后端

Low-latency ASR using SpeechBrain StreamingASR and torchaudio StreamReader.

In this repository, I try to combine k2 with speechbrain to decode well and fastly.

VoxLingua107 recipe for SpeechBrain

Target speaker automatic speech recognition (TS-ASR)

SpeechBrain中文文档

No description available

audio, NLP, ML with huggingface, nvidia/nemo, speechbrain

A recipe for disfluency detection on the LibriStutter dataset using SpeechBrain

No description available

Incremental learning for automatic speech recognition (ASR)

Processing EEG data using Speechbrain-MOABB and model tuning to get best results

A Streamlit web app for speaker diarization and identification in audio files. Upload or record audio, transcribe conversations, and automatically segment and label speakers using reference samples. This app makes it easy to analyze multi-speaker audio, export transcripts, and identify "who spoke when" for meetings, interviews, and more.

Record voice, transcribe a prompt, picturize the prompt, create variations, get description of a celebrity and upload, other use cases on KB

Implementation of different curriculum learning (CL) methods for speechbrain's ASR recipes.

[Research] 2nd place solution at L3DAS21 challenge Task 1. Using FCN architecture and Perceptual Losses. Implemented with the SpeechBrain toolkit

Attempting to build a custom pipeline using 100k hours of Russian speech data, leveraging Wav2Vec2 and speechbrain/spkrec-ecapa-voxceleb for embedding extraction. This will involve employing a combination of non-standard clustering approaches.

Train custom wake word models with openWakeWord. A granular 13-step pipeline with compatibility patches for torchaudio 2.10+, Piper TTS, and speechbrain. Generates tiny ONNX models (~200 KB) for real-time keyword detection — like building your own "Hey Siri" trigger. WSL2/Linux + CUDA required.

Audio source separation model with a Whisper ECAPA-TDNN counter and pre‑trained speechbrain/sepformer-libri3mix and speechbrain/sepformer-wsj02mix for speech separation, implemented with SpeechBrain.