Update README benchmarks section with full comparison and CPU speed tables

README.md

@@ -134,15 +134,49 @@ for chunk in tts.synthesize_stream("Một đoạn văn bản dài hơn…", voic
 ## Benchmarks
 
 Measured on **[ZeroBench-TTS](https://huggingface.co/datasets/zeroweight-ai/ZeroBench-TTS)**
-against OmniVoice, XTTS-v2-vietnamse and viXTTS, on normalized text (dates,
-numbers and acronyms already spoken out): **0.56 % WER** (4× fewer errors than
-the next-best system, median 0.00 % on all four subsets), **2.91 UTMOS**, and
-**0.029 s** of dead air. On CPU it runs ~2× faster than real time (RTF 0.5×)
-with first audio in ~70 ms of streaming.
+
+Every system reads **normalized text** — dates, numbers and acronyms already
+spoken out, from the benchmark's own curated reading. 
+
+| | **ZeroTTS** | OmniVoice | XTTS-v2-vietnamse | viXTTS |
+|---|:-:|:-:|:-:|:-:|
+| **WER** ↓ | **0.56 %** | 2.12 % | 7.27 % | 8.61 % |
+| **Naturalness** (UTMOS) ↑ | **2.91** | 2.75 | 2.49 | 2.34 |
+| **Voice similarity** (SSIM) ↑ | 0.938 | **0.951** | 0.941 | 0.935 |
+| **Dead air** (excess silence) ↓ | **0.029 s** | 0.386 s | 0.568 s | 0.215 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 |
+| Size | **81 M**, CPU | 3.1 GB, GPU | 1.9 GB, GPU | 1.9 GB, GPU |
+
+**4× fewer word errors than the next-best system**, and the only one of the four
+that's actually fast on CPU — the others 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)**
-on GitHub.
+
+### 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