本文为原文前 6,000 字符的节选翻译，完整内容请查看原文。

A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.

Back to Articles A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks. Published November 5, 2026.

回到文章：深入了解 Phonon TTS。推出在基准测试中名列前茅的高质量端侧语音技术。发布于 2026 年 11 月 5 日。

Two weeks ago, we opened up Phonon, our small text-to-speech model that runs on your own device. It needs no server and no network, and what you type stays with you. We’ve been very happy with the reception so far. So here’s the story of the launch: why we built Phonon, what it can do, the demos that came with it, and how it did on Hugging Face’s Open TTS Leaderboard.

两周前，我们开源了 Phonon，这是一个可以在您自己的设备上运行的小型文本转语音模型。它不需要服务器，也不需要网络，您输入的内容将保留在本地。到目前为止，我们对市场反馈感到非常满意。因此，这里是关于此次发布的故事：我们为何构建 Phonon，它能做什么，随之而来的演示案例，以及它在 Hugging Face 的 Open TTS 排行榜上的表现。

Why a voice should live on your device. Most voice AI lives in the cloud: every sentence makes a round trip to a server. Phonon runs on your own device instead. Your text stays there and the voice works without a network. We’ve also opened the weights. 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, so they’re free for non-commercial projects. If you’re building something commercial with those, pricing is based on your number of users, not how much speech you generate.

为什么语音应该驻留在您的设备上？大多数语音人工智能都运行在云端：每一句话都需要往返于服务器。而 Phonon 则运行在您自己的设备上。您的文本保留在本地，语音功能无需网络即可工作。我们还公开了权重。Phonon 31m 采用 Apache 2.0 协议，因此您可以免费使用它，甚至可以用于商业产品。其他模型采用 CC BY-NC 协议，因此可免费用于非商业项目。如果您使用这些模型构建商业产品，定价将基于您的用户数量，而不是您生成的语音量。

Getting a good voice to run on a small device has always been hard. You could make the model small, but then it didn’t sound great. Or you could make it sound great, but then it wouldn’t fit. We addressed both. We announced Phonon in April and spent the next six months building on it with partners and early users. Their feedback helped us to shape this release. The models got smaller, and we added more languages and support for custom voices through voice cloning and voice design.

让高质量的语音在小型设备上运行一直很困难。您可以缩小模型，但音质会变差；或者您可以追求音质，但模型会变得过大。我们解决了这两个问题。我们在四月份宣布了 Phonon，并在接下来的六个月里与合作伙伴及早期用户共同完善它。他们的反馈帮助我们塑造了此次发布版本。模型变得更小了，我们还增加了更多语言，并通过语音克隆和语音设计支持了自定义语音。

Phonon is based on the Pocket TTS lineage and as such uses flow matching to generate the audio which is very different from the modeling techniques used in other TTS systems. This is what allowed us to bring this level of quality to such a small model but it created new challenges while training the models. The training was initially numerically unstable and required several tricks to produce quality results on all five languages.

Phonon 基于 Pocket TTS 系列，因此使用流匹配（flow matching）来生成音频，这与其他 TTS 系统使用的建模技术截然不同。这使我们能够在如此小的模型中实现这种质量水平，但也给模型训练带来了新的挑战。训练初期在数值上是不稳定的，需要一些技巧才能在所有五种语言上产生高质量的结果。

What we released. Phonon comes in two sizes, at about 65 MB and 75 MB. Together they speak English, French, German, Spanish and Portuguese. There are more than 100 built-in voices, and you can create your own in Gradium Studio. Phonon also handles the important things that most other TTS systems often stumble on, like phone numbers, dates and codes. It runs on phones, laptops, single-board computers and in a browser tab, so you can hear it right now.

我们发布的内容：Phonon 有两种尺寸，分别为约 65 MB 和 75 MB。它们共同支持英语、法语、德语、西班牙语和葡萄牙语。内置了 100 多种语音，您还可以在 Gradium Studio 中创建自己的语音。Phonon 还能处理大多数其他 TTS 系统经常出错的重要内容，例如电话号码、日期和代码。它可以在手机、笔记本电脑、单板计算机和浏览器标签页中运行，因此您现在就可以试听。

Phonon 31m speaks English and is Apache 2.0, so you can build on it freely, including for commercial products. Phonon 40m covers all five languages under a non-commercial license.

Phonon 31m 支持英语并采用 Apache 2.0 协议，因此您可以自由地在其基础上进行构建，包括用于商业产品。Phonon 40m 涵盖所有五种语言，采用非商业许可协议。

Demos and first projects. The launch came with demos from friends and from our own 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 and keeps you going. 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’s also a sci-fi game voiced on the player’s own machine, and Proust read aloud page by page.

演示与首批项目：此次发布附带了来自朋友和我们团队的演示。在关闭 Wi-Fi 的 Apple Watch 上，Phonon 可以一步步指导您冲泡咖啡。在没有信号的 iPhone 上，它可以为您计数并鼓励您坚持下去。来自 Autonomous 的机器人 Lamp 戴着分院帽，并以角色身份进行对话。在浏览器中，离线的小恐龙跑酷游戏变成了一个由 Rex 主持的广播节目。此外还有在玩家本地机器上配音的科幻游戏，以及逐页朗读的普鲁斯特作品。

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 it’s still trending on Hugging Face.

Hugging Face 团队构建了 Transformers 集成，因此开发者从第一天起就可以在 Transformers 中使用 Phonon。开发者们也已经开始基于它进行开发。ExecuTorch 转换版本可以在浏览器的访客 GPU 上运行 Phonon，Micdrop 将其引入了 Node.js 应用程序，而像 smallest-s2s 这样的 Spaces 项目也基于它构建。Phonon 在发布的前两周内下载量超过 20 万次，并且在 Hugging Face 上依然保持热门趋势。

How Phonon did on the Hugging Face Open TTS Leaderboard. The Open TTS Leaderboard is a very new project from Hugging Face, and we love it. With more than 9,000 TTS models on the Hugging Face Hub, the leaderboard gives developers one place to compare open models and hear them side by side. Across 23 languages, models are evaluated on intelligibility (word error rate of the outputs), speed (inverse real-time factor and time-to-first audio), and naturalness (UTMOS for English and KOMOS for Korean).

Phonon 在 Hugging Face Open TTS 排行榜上的表现：Open TTS 排行榜是 Hugging Face 的一个非常新的项目，我们非常喜欢它。Hugging Face Hub 上有超过 9000 个 TTS 模型，该排行榜为开发者提供了一个对比开源模型并进行试听的平台。在 23 种语言中，模型根据可懂度（输出的词错误率）、速度（逆实时因子和首个音频生成时间）以及自然度（英语使用 UTMOS，韩语使用 KOMOS）进行评估。

Fixed-voice and voice cloning are evaluated, with an additional speaker similarity metric for the latter. The leaderboard also has a Listen tab, where you can conveniently compare TTS outputs yourself and vote for the ones you like the most. On intelligibility, Phonon ranks first in German and Spanish, and in the top three in every language we evaluated, next to models up to 100 times its size. Speed is where we see Phonon truly shine, as the models top the CPU streaming benchmark on time-to-first audio.

排行榜评估了固定语音和语音克隆，后者还增加了说话人相似度指标。排行榜还有一个“试听”选项卡，您可以方便地亲自比较 TTS 输出并为您最喜欢的模型投票。在可懂度方面，Phonon 在德语和西班牙语中排名第一，在我们评估的每种语言中均位列前三，与比其大 100 倍的模型并驾齐驱。速度是 Phonon 真正闪光的地方，因为这些模型在 CPU 流式传输基准测试的首个音频生成时间上名列前茅。

Inference. We built a new inference framework, ptts, specifically for Phonon. The goal of ptts is to get the best performance on a wide variety of devices and as such 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 🤗 Transformers, from v5.21, as its reference PyTorch implementation.

推理：我们专门为 Phonon 构建了一个新的推理框架 ptts。ptts 的目标是在各种设备上获得最佳性能，因此支持 10 多种后端：CPU（Intel、Arm、WebAssembly）、GPU（CUDA、Vulkan、Metal、WebGPU）和 NPU（Apple Core ML、Snapdragon®）。从 v5.21 版本开始，Phonon 也作为参考 PyTorch 实现集成在 🤗 Transformers 中。

Where ptts runs a single stream on device, Transformers targets GPU throughput: batched generation, full-graph torch.compile, e.g., for synthesizing speech datasets at scale. Being native PyTorch, it is a base to build upon, whether to integrate Phonon into your own stack or to study the model itself. It belongs to the Pocket TTS lineage, so we first added Pocket TTS to Transformers. Phonon sits on top as a modular model: about 200 lines describing only what differs. Phonon generates audio latents autoregressively: each step outputs a latent that is fed back as the next input, and the training loss is computed on them. Latents are the model’s inputs, outputs and targets, as text tokens are for a language model, so the codec mapping audio to latents plays the role of a tokenizer. Transformers ships it as such: an audio tokenizer, held by the processor next to the text tokenizer.

当 ptts 在设备上运行单流时，Transformers 则侧重于 GPU 吞吐量：例如用于大规模合成语音数据集的批量生成和全图 torch.compile。作为原生 PyTorch 实现，它是进一步构建的基础，无论是将 Phonon 集成到您自己的技术栈中，还是研究模型本身。它属于 Pocket TTS 系列，因此我们首先将 Pocket TTS 添加到了 Transformers 中。Phonon 作为模块化模型位于其上：仅用约 200 行代码描述差异部分。Phonon 以自回归方式生成音频潜变量：每一步输出一个潜变量并将其反馈作为下一个输入，训练损失即基于此计算。潜变量是模型的输入、输出和目标，就像文本标记对于语言模型一样，因此将音频映射到潜变量的编解码器扮演了分词器的角色。Transformers 将其作为音频分词器发布，由处理器与文本分词器一起持有。