Kana & Mari's SoundRepos (English)

This repository is a Qt-based desktop application for generating short-form video content from a user-provided topic. It uses AI to create voiceover scripts and titles, then combines them with video素材 from Pexels/Pixabay and configurable voice/video settings to produce ready-to-share promotional videos.

This repository provides a C++/GGML implementation of Tortoise-TTS, a text-to-speech system that generates speech from a text prompt and a selected voice sample. It includes command-line tooling to run inference, load voice/model binaries, and build for CPU, CUDA, and experimental Metal support.

This repository is a PyTorch implementation of ByteDance’s Cross-Speaker Emotion Transfer model for text-to-speech. It supports multi-speaker, multi-emotion speech synthesis, including training, preprocessing, inference, batch synthesis, TensorBoard logging, and optional use of external speaker embeddings and vocoders.

CosyVoice2-Ex is an extension of the CosyVoice text-to-speech project that adds features such as pretrained voices, 3-second voice cloning, natural-language control, automatic speech recognition, voice saving, and an API. It provides both a Gradio web UI and a Flask-based HTTP API for generating speech on Windows, Linux, and macOS.

This repository provides a Home Assistant custom integration that adds text-to-speech announcements using OpenAI’s TTS API and other OpenAI-compatible backends. It supports multiple TTS profiles, voice/model validation, audio format selection, chimes, loudness normalization, volume restore, pause/resume playback, and multi-speaker targeting.

SyntaSpeech is the official PyTorch implementation of an IJCAI 2022 paper on syntax-aware, non-autoregressive text-to-speech synthesis. It extends PortaSpeech with syntactic graph construction/encoding and multi-length adversarial training, and supports LJSpeech, Biaobei, and LibriTTS workflows for preprocessing, training, and inference.

This repository is a PyTorch implementation of Tacotron, an end-to-end text-to-speech synthesis model. It provides code for text preprocessing, dataset loading, model definition (encoder, attention-based decoder, and post-processing network), training, and waveform synthesis from text.

DramaClaw is a source-available, self-hostable industrial drama-production pipeline for turning manuscripts into short dramas. It covers the full workflow from novel parsing and episode planning to script generation, storyboards, voice synthesis, video composition, and export, with both a main pipeline and a freeform canvas for exploration.

This repository provides sample code, notebooks, and pretrained checkpoints for training and testing Persian/Farsi text-to-speech models using Coqui TTS. It focuses on model recipes for architectures such as GlowTTS, VITS, Tacotron2, HiFi-GAN, and WaveRNN, with examples for single-speaker and multispeaker training, fine-tuning, and inference.

AzSpeech is an Unreal Engine plugin that integrates Microsoft Azure Speech Cognitive Services into the engine. It provides asynchronous speech recognition and synthesis features, plus an editor tool for generating USoundWave assets directly inside Unreal Engine.

This repository implements MelNet, a generative model for audio in the frequency domain, with support for both unconditional audio generation and text-to-speech synthesis. It includes tiered training and inference pipelines, mel-spectrogram reconstruction to waveform, dataset loaders for audio and text-conditioned data, and Korean/English text normalization utilities.

This repository provides custom Home Assistant components for playing TTS audio through a Bluetooth speaker while coordinating with Bluetooth-based presence tracking. It includes a media player integration for TTS playback and an optional replacement Bluetooth tracker that pauses scanning during playback to avoid audio disruption.

This repository is a curated collection of AI/LLM-related resources, mainly linking to models, applications, datasets, tutorials, and reviews across chat LLMs, voice conversion/TTS, stable diffusion, prompt engineering, and virtual digital human topics. It serves as a reference hub for discovering and comparing popular AI tools and projects, with many entries pointing to external demos, papers, an...

Whisplay-AI-Chatbot is a pocket-sized AI chatbot device project for Raspberry Pi Zero 2W / Pi 5 that combines push-to-talk voice interaction, on-device display/LED feedback, and support for multiple AI backends. It includes features such as ASR, LLM, TTS, wake word detection, image generation, RAG/knowledge indexing, plugins, and optional offline/local AI services and accelerators.

TensorVox is a desktop neural text-to-speech application built with C++ and Qt. It loads and runs multiple TTS model families, with modern support centered on ONNX Runtime/DirectML and legacy compatibility for TensorFlow and LibTorch models, plus features like batch denoising, spectrogram/attention visualization, and audio export/playback.

Vox Box is a Python server that provides OpenAI-compatible text-to-speech and speech-to-text APIs. It supports multiple backend models and frameworks, including Whisper/Faster-Whisper, FunASR, Bark, CosyVoice, and Dia, and exposes endpoints for speech generation, transcription, model listing, voices, and health checks.

This repository implements CoMoSpeech, a consistency-model-based text-to-speech and singing voice synthesis system designed to generate audio in a single sampling step. It includes training and inference scripts, dataset loading, and HiFi-GAN vocoder integration for waveform generation, with support for both teacher and distilled student models.

This repository is an Arduino text-to-speech library that converts text into spoken output using precomputed English vocabulary/phoneme rules and embedded sound data. It supports multiple Arduino-compatible boards and microcontrollers with PWM or DAC output, and requires an external amplifier for audio playback.

htgo-tts is a Go text-to-speech package that generates MP3 speech audio from text using Google Translate TTS endpoints. It can save speech files, play them back through either an external mplayer process or a native audio handler, and supports multiple languages plus optional proxy configuration.

BreezeAPP 是一款為 Android 和 iOS 平台開發的純手機 AI 應用程式。從 App Store下載，即可在不連網的狀態下享受多項 AI 功能。源碼由聯發創新基地(MediaTek Research)提供。我們旨在推廣兩個概念: 人人都可以在自己的手機上自由選擇並運行不同的LLM - one is free to choose one's own LLM to run on a phone，以及任何app開發者都可以輕鬆寫作創意的純手機AI應用 - any dev can create purely phone-based AI apps easily。