Phonon TTS: High-Quality, On-Device Speech That Tops the Benchmarks

Hugging Face launched Phonon, a small text-to-speech (TTS) model that runs directly on your device, eliminating the need for a server or network connection. This means your text stays private and the voice works offline. Phonon comes in two sizes, 65 MB and 75 MB, and supports English, French, German, Spanish, and Portuguese with over 100 built-in voices.

Phonon Runs On-Device With No Network Required

Most voice AI lives in the cloud, sending every sentence to a server and back. Phonon runs on your own device instead. Your text stays there and the voice works without a network. The weights are open: Phonon 31m is Apache 2.0, so you can use it for free, even in a commercial product. The other models are CC BY-NC, free for non-commercial projects. If you are building something commercial with those, pricing is based on your number of users, not how much speech you generate.

Phonon is based on the Pocket TTS lineage and uses flow matching to generate audio, a technique that differs from other TTS systems. This allowed the team to bring high quality to a small model, but the training was initially numerically unstable and required several tricks to produce quality results across all five languages.

Benchmark Performance: First in German and Spanish

Phonon made a significant impact on the Hugging Face Open TTS Leaderboard, which evaluates models across 23 languages on intelligibility (word error rate), speed (inverse real-time factor and time-to-first audio), and naturalness (UTMOS for English and KOMOS for Korean). On intelligibility, Phonon ranks first in German and Spanish, and in the top three in every language it was evaluated in, next to models up to 100 times its size. Speed is where Phonon truly shines, as the models top the CPU streaming benchmark on time-to-first audio.

The team built a new inference framework, ptts, specifically for Phonon. It supports more than 10 backends: CPU (Intel, Arm, WebAssembly), GPU (CUDA, Vulkan, Metal, WebGPU), and NPU (Apple Core ML, Snapdragon). Phonon is also available in the Transformers library from v5.21, as its reference PyTorch implementation. Where ptts runs a single stream on device, Transformers targets GPU throughput for batched generation and synthesizing speech datasets at scale.

Launch Demos and Developer Adoption

The launch came with demos from friends and the team. On an Apple Watch with Wi-Fi off, Phonon talks you through a coffee brew step by step. On an iPhone with no signal, it counts your reps during a workout. A robot called Lamp, from Autonomous, wears a sorting hat and talks in character. In a browser, an offline dino runner becomes a radio show hosted by Rex. There is also a sci-fi game voiced on the player's own machine, and Proust read aloud page by page.

The Hugging Face team built the Transformers integration, so developers could use Phonon in Transformers from day one. Developers have started building on it too. An ExecuTorch conversion runs Phonon on a visitor's GPU in the browser, Micdrop brings it to Node.js apps, and Spaces like smallest-s2s build on it. Phonon was downloaded more than 200,000 times in its first two weeks and is still trending on Hugging Face.

What You Can Do Today

To try Phonon TTS, you can download it from the Hugging Face Hub and run it on your device. The model is available in the Transformers library, making it easy to integrate into your own projects. You can also explore the demos provided by Hugging Face. For developers, Phonon TTS offers a versatile tool for building applications that require high-quality, offline speech synthesis. You can also create custom voices using Gradium Studio, allowing you to tailor the model to your specific needs.