Open source. Open possibilities.

ESPnet is an end-to-end speech processing toolkit built on PyTorch, covering ASR, TTS, speech translation, enhancement, diarization, and more, with reproducible recipes and pretrained models.

WhisperLiveKit is a self-hosted, ultra-low-latency speech-to-text pipeline. It provides real-time transcription, speaker diarization, and translation via WebSocket and OpenAI/Deepgram-compatible APIs.

Stable Baselines3 (SB3) is a set of reliable PyTorch implementations of reinforcement learning algorithms designed for research and industry replication.

Made With ML is an open-source course and codebase teaching how to design, develop, deploy and iterate on production-grade machine learning applications, covering MLOps, testing, serving, CI/CD and continual learning.

A comprehensive toolkit for computer vision that provides model-agnostic building blocks for data loading, visualization, and dataset management.

Cog packages machine learning models into production-ready Docker containers, automatically handling CUDA compatibility and generating an HTTP inference API from Python type definitions.

Pyro is a flexible, scalable deep probabilistic programming library built on PyTorch, supporting universal probabilistic models and automated inference.

timm is a large collection of PyTorch image models, layers, optimizers, schedulers, data loaders and training/eval scripts, offering many pretrained backbones such as ResNet, EfficientNet, ViT, ConvNeXt and Swin.

EasyOCR is a ready-to-use optical character recognition library supporting 80+ languages and major writing scripts. It offers a simple Python API and CLI, with PyTorch-based detection and recognition models, plus options for custom training.

Premove ITN is an open-source, context-aware inverse text normalization tool for English voice-agent transcripts. It converts spoken ASR output into canonical written forms using a generate-score-decode pipeline with DeBERTa contextual scoring.

InsightFace is a state-of-the-art open-source 2D and 3D face analysis project providing face recognition, detection, alignment, swapping, and liveness detection with pretrained models, a Python package, and a self-hosted server.

einops is a flexible tensor operations library using Einstein-like notation for readable, reliable code across PyTorch, JAX, TensorFlow, NumPy and more.

Hugging Face Transformers is a model-definition framework for machine learning across text, vision, audio, and multimodal tasks. It provides a unified API for inference and training with over 1 million pretrained checkpoints.

SenseVoiceSmall is an open-source speech foundation model for ASR, language ID, emotion recognition and audio event detection, covering Mandarin, Cantonese, English, Japanese and Korean with fast non-autoregressive inference.