Run Audio8 TTS Locally — 0.6B & 0.1B Models Tested

Run Audio8 TTS Preview 0.6B and the smaller 0.1B model locally, and test AI voice generation and zero-shot voice cloning using reference audio. In this video, I show you how to install and run Audio8 TTS locally, step by step. We start by installing the required tools, cloning the Audio8 repository, creating and activating the Python environment, installing CUDA-enabled PyTorch, downloading the Audio8 TTS Preview 0.6B model, and verifying that PyTorch can access the GPU. Then, we use a local playground UI to test text-to-speech and voice generation with reference audio. After testing the 0.6B model, we also test the much smaller Audio8 TTS Preview 0.1B model. This compact version supports speech generation and zero-shot voice cloning, allowing us to provide reference audio and generate new speech based on that voice. The 0.1B model is particularly interesting because its main generative model is approximately 170 million parameters, while the neural audio codec decoder is approximately 120 million parameters. Together, the complete audio generation stack remains relatively compact compared with many modern multilingual TTS systems. For a fair comparison, I use the same reference audio, reference text, and test script with both models, then compare the results side by side. I used an RTX 5090 for this demonstration. However, you do not need an RTX 5090 specifically to install and experiment with Audio8. You can test it on other consumer hardware as well, although generation speed and performance will depend on your system. In this video: Audio8 TTS Preview 0.6B Audio8 TTS Preview 0.1B Local TTS installation Text-to-speech generation Zero-shot voice cloning Reference-audio voice generation 0.6B vs 0.1B comparison Local GPU inference RTX 5090 testing Step-by-Step Commands Step 1 — Install Git winget install --id Git.Git -e Step 2 — Install Hugging Face CLI pip install -U "huggingface_hub[cli]" Step 3 — Clone the Audio8 Repository cd /d D:\Addis_Pulse_Projects\audio git clone https://github.com/Audio8-AI/Audio8_T... Audio8-TTS cd Audio8-TTS Step 4 — Install uv winget install --id astral-sh.uv -e Step 5 — Create the Python Environment uv venv --python 3.12 .venv Step 6 — Activate the Environment .venv\Scripts\activate Step 7 — Install CUDA PyTorch pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 Step 8 — Install Audio8 Dependencies pip install -r requirements.txt Step 9 — Download Audio8 TTS Preview 0.6B hf download Audio8/Audio8-TTS-Preview-0.6b --local-dir model\audio8_tts_0_6B_preview Step 10 — Download Audio8 TTS Preview 0.1B hf download Edge0/Audio8-TTS-Preview-0.1b --local-dir model\audio8_tts_0_1B_preview Step 11 — Check PyTorch, CUDA, and Transformers python -c "import torch, transformers; print(torch.version, torch.cuda.is_available(), transformers.version)" Expected output: 2.x.x+cu128 True 4.x.x Step 12 — Confirm GPU nvidia-smi Chapters 00:00 Introduction 00:46 What is Audio8-TTS 02:26 Step-by-Step Commands 05:00 Run Audio8 0.6B for the First Time 06:42 Test Audio8 0.1B Model 08:14 Conclusion and Final Thoughts #Audio8TTS #texttospeech #voicecloning #localai