--- license: mit language: - vi library_name: gguf pipeline_tag: text-to-speech tags: - text-to-speech - tts - vietnamese - gguf - ggml - llama-cpp - quantized - webassembly - zero-shot - speech-synthesis - voice-cloning - vietnamese-tts - tieng-viet metrics: - wer model-index: - name: ZeroTTS results: - task: type: text-to-speech name: Zero-Shot Text-to-Speech dataset: type: zeroweight-ai/ZeroBench-TTS name: ZeroBench-TTS split: test metrics: - type: wer value: 1.03 name: WER (%) — raw text - type: utmos value: 2.91 name: UTMOSv2 naturalness MOS - type: speaker_similarity value: 0.936 name: Speaker similarity (WavLM-SV cosine) - type: excess_silence value: 0.029 name: Excess silence (s) - task: type: text-to-speech name: Zero-Shot TTS — monolingual Vietnamese dataset: type: zeroweight-ai/ZeroBench-TTS name: ZeroBench-TTS (vietnamese) config: vietnamese split: test metrics: - type: wer value: 0.16 name: WER (%) — raw text - task: type: text-to-speech name: Zero-Shot TTS — Vietnamese/English code-switching dataset: type: zeroweight-ai/ZeroBench-TTS name: ZeroBench-TTS (code_switch) config: code_switch split: test metrics: - type: wer value: 0.97 name: WER (%) — raw text - task: type: text-to-speech name: Zero-Shot TTS — cross-lingual voice prompt dataset: type: zeroweight-ai/ZeroBench-TTS name: ZeroBench-TTS (cross_lingual) config: cross_lingual split: test metrics: - type: wer value: 1.42 name: WER (%) — raw text - task: type: text-to-speech name: Zero-Shot TTS — acronyms, dates, numbers dataset: type: zeroweight-ai/ZeroBench-TTS name: ZeroBench-TTS (challenging) config: challenging split: test metrics: - type: wer value: 1.75 name: WER (%) — raw text --- ZeroTTS — Vietnamese zero-shot text-to-speech # ZeroTTS — GGUF ### The ZeroTTS weights packaged as GGUF, for the [ggml](https://github.com/ggml-org/ggml) runtime instead of onnxruntime. Same model, ~4× smaller download, and it runs in the browser under WebAssembly. This is a **repackaging**, not a different model. The weights are the same ones in [`zeroweight-ai/ZeroTTS`](https://huggingface.co/zeroweight-ai/ZeroTTS); the `f32` file here is verified to produce **bit-identical frame codes** to the ONNX graphs for the same text, voice and random draws. If you want the ONNX runtime or the `zerotts` pip package, use that repo — this one is for the ggml/WASM runtime in [`cpp/`](https://github.com/zeroweight-ai/ZeroTTS/tree/main/cpp). **The most accurate open Vietnamese TTS we know of — 4× fewer word errors than the next best model**, and it runs faster than real time on a laptop CPU. * 🎯 **Ultra-natural** — 2.91 UTMOS above every other open Vietnamese system, with near-zero dead air (0.029 s). * 🗣️ **Zero-shot voice cloning** — a voice is a small latent array; drop it in and the model speaks in it, cloned from as little as 3 seconds of reference audio (up to 30 seconds). No fine-tuning, no per-speaker training. * ⚡ **Real-time on CPU, streaming** — ~2× faster than real time (RTF 0.5×), first audio chunk in ~70 ms. No GPU required. * 🇻🇳 **Built for Vietnamese** — tones, code-switched English, and reads `31/12/2025` and `ZeroTTS` without text normalizer. * Code, examples, browser demo: **https://github.com/zeroweight-ai/ZeroTTS** * Benchmark dataset: **https://huggingface.co/datasets/zeroweight-ai/ZeroBench-TTS** * Blogpost: **https://zeroweight.ai/blog/zero-tts** ## Samples **Two-speaker conversation** **Long-form narration** **News read, code-switched English** **Cross-lingual** Reference audio (Vietnamese) Output (English) ## Files | path | what it is | |---|---| | `gguf/zerotts-q8_0.gguf` | **the default** — 206 MB, the text encoder, global decoder and depth decoder | | `gguf/zerotts-q4_0.gguf` | 124 MB, when the download budget is tight | | `gguf/zerotts-f32.gguf` | 772 MB, unquantized; the reference the others are measured against | | `onnx/codec/` | the MOSS audio-tokenizer **decoder**, still ONNX (Apache-2.0, ~45 MB) — it is not on the per-frame hot path | | `voices/` | the voice latents; a voice is a 10×768 float32 array | | `tokenizer.json`, `config.json`, `null_voice_emb.npy`, `silence_frame.npy` | runtime side files, identical to the ONNX repo | You need **one** `.gguf` plus `onnx/codec/`, `voices/` and the side files. ### Which one to take Measured in Chrome on a 14-core Apple-silicon laptop, one 3.2 s utterance, frame generation only. "drift" is the share of sampled codes that come out differently from `f32` when both are driven along the same code sequence — the `` stop decision was identical in every build. | file | size | drift vs f32 | 1 thread | 4 threads | 8 threads | |---|---:|---:|---:|---:|---:| | `f32` | 772 MB | — | **2.40×** | **6.1×** | **6.2×** | | `q8_0` | 206 MB | 4.9% | 1.06× | 3.6× | 5.3× | | `q4_0` | 124 MB | 36.5% | 1.27× | 4.1× | 5.8× | For reference, the ONNX graphs in [`zeroweight-ai/ZeroTTS`](https://huggingface.co/zeroweight-ai/ZeroTTS) run at 4.2× on the same machine, at the 4 threads onnxruntime-web caps itself to. **Quantization here buys download size, not speed.** WebAssembly's SIMD has no integer dot-product instruction, so a `q8_0` block costs roughly 20 SIMD ops per 32 multiply-accumulates where `f32` costs 8 fused multiply-adds — on a machine with memory bandwidth to spare the extra instructions cost more than the saved bytes. Take `q8_0` unless you have a reason not to: it is 4× smaller than the ONNX graphs and faster than them at the thread counts each actually uses (5.3× on 8 threads against 4.2× on 4). Take `f32` if you want bit-exactness, the highest speed at any thread count, or you are running on very few cores — it is the fastest build here, just the largest. `q4_0` changes a third of the drawn codes, which is audible as a different take rather than a broken one, but it is the wrong trade unless 124 MB is the binding constraint. Numbers on hardware that is genuinely short of *bandwidth* rather than compute should look different; the benchmark page in the repo re-runs this table in one click. ## Usage The runtime is [`cpp/`](https://github.com/zeroweight-ai/ZeroTTS/tree/main/cpp) in the main repository — a ggml implementation of the generation loop, building natively and to WebAssembly. ```bash git clone --recurse-submodules https://github.com/zeroweight-ai/ZeroTTS cd ZeroTTS/cpp cmake -B build -DCMAKE_BUILD_TYPE=Release . && cmake --build build -j ./build/zerotts-bench /path/to/zerotts-q8_0.gguf voice.bin text_ids.txt ``` In the browser, the demo in [`js/`](https://github.com/zeroweight-ai/ZeroTTS/tree/main/js) uses this repo by default: ```bash cd ZeroTTS/js && npm install && npm run dev ``` `voice.bin` is the raw little-endian float32 blob under `voices//` — `voice.npz` is the same latents for numpy callers. For the Python package, the ONNX graphs and `pip install zerotts`, use [`zeroweight-ai/ZeroTTS`](https://huggingface.co/zeroweight-ai/ZeroTTS) instead. ## Benchmarks These are the unquantized model's quality numbers, measured through the PyTorch runtime. `f32` here is that model exactly; see the drift column above for what the quantized files change. Measured on **[ZeroBench-TTS](https://huggingface.co/datasets/zeroweight-ai/ZeroBench-TTS)** Every system reads **raw text** — dates, numbers and acronyms verbatim, exactly as they appear in the wild, with no text frontend in front of the model. | | **ZeroTTS** | OmniVoice | XTTS-v2-vietnamse | viXTTS | |---|:-:|:-:|:-:|:-:| | **WER** ↓ | **1.03 %** | 4.13 % | 16.42 % | 18.40 % | | **Naturalness** (UTMOS) ↑ | **2.91** | 2.76 | 2.43 | 2.35 | | **Voice similarity** (SSIM) ↑ | 0.936 | **0.950** | 0.940 | 0.935 | | **Dead air** (excess silence) ↓ | **0.029 s** | 0.340 s | 0.532 s | 0.233 s | | **RTF, CPU** ↓ | **0.50×** | 6.12× | 0.71× | 0.73× | | **Time to first audio, CPU** ↓ | **~70 ms** | ~34 s | ~6.1 s | ~5.1 s | | **Parameters** ↓ | **202 M** | 775 M | 467 M | 467 M | **4× fewer word errors than the next-best system**, and the fastest of the four on CPU. The gap is much wider in latency than in throughput: the two XTTS fine-tunes also beat real time (0.71×) but need seconds to emit their first sample, while OmniVoice is 6× *slower* than real time. All three are sized and tuned for a GPU, and it shows. Full comparison tables, per-subset breakdowns, and CPU speed methodology: **[docs/BENCHMARKS.md](https://github.com/zeroweight-ai/ZeroTTS/blob/main/docs/BENCHMARKS.md)** ### Speed — CPU RTF (realtime factor, wall-clock synthesis time ÷ output audio duration — lower is faster; below 1× is faster than real time) and time-to-first-audio, all measured **on CPU**, single request, 8 inference threads pinned to a dedicated core pool (no other synthesis running concurrently). Three Vietnamese samples — short (26 chars), medium (77 chars), long (227 chars) — each run 6 times with the first 2 (cold-cache) discarded; figures below are the mean of the remaining 4. | | **ZeroTTS** | OmniVoice | XTTS-v2-vietnamse | viXTTS | |---|:-:|:-:|:-:|:-:| | RTF — short | **0.51×** | 10.87× | 0.70× | 0.71× | | RTF — medium | **0.47×** | 4.82× | 0.70× | 0.70× | | RTF — long | **0.53×** | 2.67× | 0.71× | 0.78× | | TTFA — short | **53 ms** | 21.7 s | 4.02 s | 2.45 s | | TTFA — medium | **66 ms** | 28.9 s | 4.02 s | 3.72 s | | TTFA — long | **89 ms** | 52.3 s | 10.3 s | 9.22 s | ZeroTTS's time-to-first-audio comes from its real streaming path — first audio frame, not first full utterance. The three baselines have no working CPU streaming path, so their TTFA is the time to the complete utterance. ## Voices, and voice cloning A voice is a small array of speaker latents, `(1, n_voice_queries, d_model)`, shipped as a `.npz` under `voices/`. That array is the entire speaker conditioning — no reference transcript, no audio prompt. > **Voice cloning is not available in this release.** Those latents come from a > voice encoder that reads a reference clip, and that encoder is not published. > This repository ships ready-to-use voices; it cannot create new ones from > audio. > > To get latents for your own speaker, see **[zeroweight.ai](https://zeroweight.ai)** > or get in touch. Because a voice is just an array, latents obtained that way drop into `voices//voice.npz` and work with no code change. ## Intended use and limitations Built for **Vietnamese**. It handles English words embedded in Vietnamese text (`code_switch`), but it is not an English TTS system and is not evaluated as one. Do not use it to impersonate a real person, to generate speech attributed to someone without their consent, or to produce audio intended to deceive. The shipped voices are for evaluation and demos. Synthetic speech should be disclosed as synthetic wherever a listener might reasonably assume otherwise. ## Credits Speech codec: **MOSS-Audio-Tokenizer-Nano** by the OpenMOSS team, Apache-2.0. Its ONNX **decoder** graphs are redistributed under `onnx/codec/` so ZeroTTS has no external runtime dependency; the encoder is not included. See `onnx/codec/LICENSE-Apache-2.0.txt`. ```bibtex @misc{gong2026mossaudiotokenizerscalingaudiotokenizers, title={MOSS-Audio-Tokenizer: Scaling Audio Tokenizers for Future Audio Foundation Models}, author={Yitian Gong and Kuangwei Chen and Zhaoye Fei and Xiaogui Yang and Ke Chen and Yang Wang and Kexin Huang and Mingshu Chen and Ruixiao Li and Qingyuan Cheng and Shimin Li and Xipeng Qiu}, year={2026}, eprint={2602.10934}, archivePrefix={arXiv}, primaryClass={cs.SD} } ``` ## License ZeroTTS weights and code: **MIT**. The ZeroBench-TTS *dataset* is CC-BY-NC-4.0 because it redistributes reference audio from VIVOS, viVoice, phoaudiobook and Emilia. That license applies to the benchmark dataset only — **not** to these weights.