SenseVoice

SenseVoice is a speech foundation model that provides multilingual recognition, emotion recognition, and audio event detection.

SenseVoice is a speech foundation model with multiple understanding capabilities, including automatic speech recognition, spoken language identification, speech emotion recognition, and audio event detection. The released SenseVoiceSmall checkpoint supports Mandarin, Cantonese, English, Japanese, and Korean, along with emotion and audio-event tags. It uses a non-autoregressive end-to-end framework for efficient inference and provides service deployment pipelines supporting multi-concurrent requests. It is designed for developers and researchers who need high-accuracy multilingual transcription and audio analysis. Users can also utilize convenient finetuning scripts to address long-tail sample issues.

SenseVoice at a glance

Best for: Software engineer
Runs on: API, Python, self-hosted, open source
Pricing: Free
Released: July 2024
Category: AI audio tool
Website: github.com/QwenAudio/SenseVoice

Key features

Automatic speech recognition
Spoken language identification
Speech emotion recognition
Audio event detection
Low-latency non-autoregressive inference
Service deployment with multi-concurrent requests
Finetuning scripts for custom scenarios

Use it for

Transcribing multilingual speech recordings
Detecting speech emotions in audio files
Identifying sound events like laughter and applause
Processing long audio with bounded memory

Alternatives to SenseVoice

VibeVoice — Process long-form audio up to 60 minutes in a single pass.
AssemblyAI — Industry-leading speech models deliver high accuracy and low latency for voice applications.
Spectromood — Evaluates speech emotion using a general-purpose decision model.
audio.cpp — Runs modern audio models locally with high-performance, native C++ execution.
Whisper ASR Webservice — Provides a REST API for multiple ASR engines including WhisperX and Faster Whisper.
WhisperX — It enables 70x faster, batched transcription with high-accuracy alignment.
Speaches — It acts as an Ollama equivalent for running local TTS and STT models.
WeNet — Achieves state-of-the-art results on public speech datasets.
Whisper Timestamped — Generates accurate word-level timestamps without extra neural networks.
WhisperLive — It provides real-time speech-to-text using multiple backends like Faster Whisper and TensorRT.
whisper.cpp — It provides high-performance, offline speech recognition on diverse hardware.
FunASR — It supports offline, streaming, and edge deployment with multiple models.

All SenseVoice alternatives