AI Dialer is a Python full-stack application for making interruptible, low-latency AI phone calls through Twilio. It streams phone audio over WebSockets, uses Deepgram for speech recognition, OpenAI or Anthropic for conversation generation, ElevenLabs or Deepgram for speech synthesis, and provides a Streamlit interface for managing calls and viewing transcripts. The project is retained as a code reference because API changes have made it obsolete.

Related: akiani/aidialer / real time AI phone call dialer / GitHub

Bournemouth Forced Aligner is a multilingual speech forced-alignment toolkit that maps transcript text to phoneme-level timestamps in audio. It provides Python and CLI interfaces, batch processing, confidence and coverage analysis, Praat TextGrid and JSON export, mel-spectrogram alignment, and optional phoneme embeddings using CUPE neural models.

Related: tabahi/bournemouth-forced-aligner / multilingual phoneme forced aligner / GitHub

ReadAware is an open-source, local-first ebook reader with an AI agent that provides context-aware assistance, persistent reading memory, annotations, offline reading, and support for formats including EPUB, MOBI, AZW3, FB2, PDF, and comics. It includes a sandboxed plugin system for extending the reader and agent with dictionaries, text-to-speech, themes, feeds, commands, and synchronization features, plus encrypted cross-device sync.

Related: ahpxex/read-aware / offline AI ebook reader with plugins / GitHub

ASMR Dubber is a local Windows and Linux application for converting Japanese or English audio and video into Chinese dubbed, bilingual, and subtitled media while preserving the original sound. It provides stepwise speech recognition, translation, voice synthesis, mixing, subtitle editing, batch processing, timestamp correction, and support for local and cloud-based AI backends.

Related: EveningStudy/asmr-dubber / Japanese English to Chinese dubbing / GitHub

pyVideoTrans is an open-source video translation and dubbing application that converts speech to subtitles, translates subtitles, synthesizes multilingual speech, and synchronizes the resulting audio and video. It supports local and online ASR, translation, and TTS providers, speaker diarization, voice cloning, interactive editing, CLI, GUI, WebUI, Docker, and GPU acceleration.

Related: jianchang512/pyvideotrans / video translation and AI dubbing tool / GitHub

A high-performance Go proxy that converts Google AI Studio web requests into OpenAI Chat and Responses, Anthropic, and Gemini-compatible APIs. It supports multi-account scheduling, streaming, Google tools, file and transcription APIs, TTS, image and video generation, Live and Robotics WebSockets, and a web-based account and service manager using Camoufox for authenticated browser automation.

Related: Mag1cFall/AIStudio2API / Google AI Studio compatible API / GitHub

TalkWithMe is a local, single-user web chat application that connects to an OpenAI-compatible llama.cpp server and supports multi-persona group conversations. It provides persistent chat rooms, persona management, optional text-to-speech and speech-to-text, MCP tool calling, persona memories, audio replay, and multiple themes.

Related: scorbo2/TalkWithMe / local multi persona AI chat with voice / GitHub

MOSS-TTS Family is an open-source suite of speech and sound generation models for high-fidelity multilingual text-to-speech, voice cloning, long-form narration, multi-speaker dialogue, real-time voice agents, voice design, and sound effects. It provides model inference examples, fine-tuning guides, browser demos, torch-free llama.cpp deployment, and integrations with SGLang-Omni and vLLM-Omni.

Related: OpenMOSS/MOSS-TTS / multilingual voice cloning TTS / GitHub

MSMC-TTS is a PyTorch implementation of a multi-stage, multi-codebook neural text-to-speech system. It trains MSMC-VQ-VAE or MSMC-VQ-GAN acoustic autoencoders and multi-stage predictors, with dataset processing, checkpointing, inference, audio generation, and multi-GPU training support.

Related: hhguo/MSMC-TTS / neural text to speech with compact codes / GitHub

DetailTTS is a multilingual zero-shot text-to-speech system that learns residual audio detail beyond VQ representations. It provides training and inference workflows for a VQ-VAE-based model, supports Chinese, English, Japanese, and Korean, and includes Colab demos, pretrained models, tokenizer training, and fine-tuning guidance.

Related: adelacvg/ttts / multilingual zero shot text to speech / GitHub