Reverb is an open-source inference codebase for speech recognition and speaker segmentation models, utilizing the WeNet framework for ASR and the Pyannote framework for speaker segmentation. It offers detailed model descriptions and allows users to download models from Hugging Face. Reverb aims to provide developers and researchers with high-quality tools for various speech processing tasks.

SafeEar is an innovative audio depth detection framework that can identify deep audio without relying on speech content. The framework employs a neural audio codec design to separate semantic and acoustic information from audio samples, using solely acoustic features (such as prosody and timbre) for depth detection, thereby safeguarding the privacy of speech content. SafeEar enhances the detector's capabilities by augmenting the codec in real-world scenarios, enabling it to recognize various forms of deep audio. Extensive experiments conducted on four benchmark datasets demonstrate that SafeEar is highly effective in detecting various deep technologies, achieving a minimum equal error rate (EER) of 2.02%. Moreover, it protects speech content in five languages from deciphering by both machine and human auditory analysis, as validated by our user studies revealing a word error rate (WER) exceeding 93.93%. Additionally, SafeEar has established a benchmark for anti-deep and anti-content recovery assessments, laying the groundwork for future research in audio privacy protection and depth detection.

Sherpa-onnx is a speech recognition and speech synthesis project based on the next-generation Kaldi, utilizing onnxruntime for inference. It supports a wide range of speech-related functionalities, including automatic speech recognition (ASR), text-to-speech (TTS), speaker recognition, speaker verification, language identification, and keyword detection. It is compatible with various platforms and operating systems, including embedded systems, Android, iOS, Raspberry Pi, RISC-V, and servers.

WhisperFusion is a product powered by WhisperLive and WhisperSpeech functionalities. It enables seamless AI conversation by integrating the Mistral large language model (LLM) into the real-time speech-to-text process. Both Whisper and LLM are optimized with the TensorRT engine to maximize performance and real-time processing capabilities. WhisperSpeech utilizes torch.compile for optimization. The product is focused on delivering an ultra-low latency AI real-time conversation experience.

NVAS3d is a project for estimating sound at any location within a scene containing multiple unknown sound sources. It achieves novel-view acoustic synthesis by using audio recordings from multiple microphones and the 3D geometry and materials of the scene.

Reverb is an open-source inference codebase for speech recognition and speaker segmentation models, utilizing the WeNet framework for ASR and the Pyannote framework for speaker segmentation. It offers detailed model descriptions and allows users to download models from Hugging Face. Reverb aims to provide developers and researchers with high-quality tools for various speech processing tasks.

SafeEar is an innovative audio depth detection framework that can identify deep audio without relying on speech content. The framework employs a neural audio codec design to separate semantic and acoustic information from audio samples, using solely acoustic features (such as prosody and timbre) for depth detection, thereby safeguarding the privacy of speech content. SafeEar enhances the detector's capabilities by augmenting the codec in real-world scenarios, enabling it to recognize various forms of deep audio. Extensive experiments conducted on four benchmark datasets demonstrate that SafeEar is highly effective in detecting various deep technologies, achieving a minimum equal error rate (EER) of 2.02%. Moreover, it protects speech content in five languages from deciphering by both machine and human auditory analysis, as validated by our user studies revealing a word error rate (WER) exceeding 93.93%. Additionally, SafeEar has established a benchmark for anti-deep and anti-content recovery assessments, laying the groundwork for future research in audio privacy protection and depth detection.

Sherpa-onnx is a speech recognition and speech synthesis project based on the next-generation Kaldi, utilizing onnxruntime for inference. It supports a wide range of speech-related functionalities, including automatic speech recognition (ASR), text-to-speech (TTS), speaker recognition, speaker verification, language identification, and keyword detection. It is compatible with various platforms and operating systems, including embedded systems, Android, iOS, Raspberry Pi, RISC-V, and servers.

WhisperFusion is a product powered by WhisperLive and WhisperSpeech functionalities. It enables seamless AI conversation by integrating the Mistral large language model (LLM) into the real-time speech-to-text process. Both Whisper and LLM are optimized with the TensorRT engine to maximize performance and real-time processing capabilities. WhisperSpeech utilizes torch.compile for optimization. The product is focused on delivering an ultra-low latency AI real-time conversation experience.

NVAS3d is a project for estimating sound at any location within a scene containing multiple unknown sound sources. It achieves novel-view acoustic synthesis by using audio recordings from multiple microphones and the 3D geometry and materials of the scene.

Enhance Speech from Adobe is a free AI audio filter that can process spoken audio to sound like it was recorded in a soundproof studio. It automatically removes background noise, adjusts volume balance, and improves audio quality. Users can upload audio files to the platform and have them optimized by AI algorithms. Enhance Speech from Adobe is suitable for broadcasting, podcasting, audio production, and other fields. The product is completely free to use.