Audio
index-tts
IndexTTS: industrial-level, controllable zero-shot text-to-speech with voice cloning
★ 24.4k starsNOASSERTION
View guide →MetaVoice-1B: a 1.2B-parameter foundation model for emotional, zero-shot text-to-speech
metavoice-src is an open-source audio project by metavoiceio (metavoiceio/metavoice-src on GitHub). It has 4,205 GitHub stars and 689 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 30 July 2024.
Confirm a GPU with at least 12 GB VRAM and Python 3.10–3.11.
Clone the repo and install with Poetry (recommended) or pip, including the pinned PyTorch build.
Bring up the UI or server service with Docker Compose for the fastest trial.
Run the interactive inference script and synthesize speech from a short reference clip.
Trained to capture rhythm and tone, not just intelligible speech.
Clone American and British voices from a 30-second reference recording.
Adapt to other languages/speakers with fine-tuning on as little as ~1 minute of data (per the project).
Designed to synthesize long passages, not only short clips.
An interactive Python inference session, plus Docker Compose services for a web UI or a server.
Training entry points for the first-stage LLM ship in the repo.
Clone the project and enter its directory.
git clone https://github.com/metavoiceio/metavoice-src.git
cd metavoice-srcInstall dependencies and the pinned PyTorch build as documented.
pipx install poetry
poetry install && poetry run pip install torch==2.2.1 torchaudio==2.2.1Start the inference session, then call tts.synthesise with your text and a reference clip.
poetry run python -i fam/llm/fast_inference.py
# then: tts.synthesise(text="Hello world", spk_ref_path="assets/bria.mp3")MetaVoice (metavoiceio; themetavoice.xyz) publishes the model and code.
The code is Apache-2.0 licensed and the weights are on Hugging Face; check the model card for any additional terms.
No — it runs locally on your GPU (12 GB+ VRAM required).
Cloud APIs trade control for convenience; MetaVoice-1B gives you a self-hosted foundation model you can fine-tune, with emotion and cloning as its headline features.
Clone the repo, `poetry install` with the pinned PyTorch, then run the interactive inference script — or `docker-compose up -d ui` for the web UI.
IndexTTS: industrial-level, controllable zero-shot text-to-speech with voice cloning
Unofficial PyTorch implementation of VALL-E, the neural codec language model for zero-shot TTS
A generative speech model built for daily dialogue — natural, conversational text-to-speech
Suno's text-prompted generative audio model: speech, music and sound effects from text prompts
Fast local speech-to-text across 16+ model families, powered by ggml.
Multilingual long-form dialogue speech with zero-shot voice cloning.
The most-starred other Audio projects in the LibreAI catalogue — the realistic shortlist if metavoice-src is not the right fit.
利用 AI 大模型和自动化工作流，根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.
Robust Speech Recognition via Large-Scale Weak Supervision
1 min voice data can also be used to train a good TTS model! (few shot voice cloning)
Clone a voice from 5 seconds of audio and generate arbitrary speech in real time
Port of OpenAI's Whisper model in C/C++
Free and open — grab the source, star the repo, and start creating.