Add Open TTS Leaderboard section — M2 ranks 8th, F2 ranks 13th (16 Nov 2026 snapshot, v1.6.3)
#2
by bezzam HF Staff - opened
README.md
CHANGED
|
@@ -190,6 +190,25 @@ recordings of the same sentences are scored the same way for reference.
|
|
| 190 |
These sets are small and the scores are a sanity check, not a benchmark. No
|
| 191 |
formal listening test has been run.
|
| 192 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 193 |
## Limitations
|
| 194 |
|
| 195 |
- English only, two fixed voices, no voice cloning.
|
|
|
|
| 190 |
These sets are small and the scores are a sanity check, not a benchmark. No
|
| 191 |
formal listening test has been run.
|
| 192 |
|
| 193 |
+
|
| 194 |
+
### Open TTS Leaderboard
|
| 195 |
+
|
| 196 |
+
Official leaderboard snapshot: **16 November 2026** (v1.6.3), English, **without voice cloning**, equal average across Seed-TTS, CV3-Eval and MiniMax; 50 ranked models (full-coverage English). Each tiny-voca voice is a separate row; **M2** currently ranks **8th by average WER** and **F2** ranks **13th**. The hosted leaderboard notes tiny-voca is CPU-only on both rows.
|
| 197 |
+
|
| 198 |
+
| Rank | Model | Params | Average WER % ↓ | UTMOS22 ↑ |
|
| 199 |
+
| ---: | :--- | ---: | ---: | ---: |
|
| 200 |
+
| 6 | phonon-31m-en | 41 M | 0.730 | 4.45 |
|
| 201 |
+
| 7 | Inflect Micro v3 | 12 M | 0.730 | 4.37 |
|
| 202 |
+
| **8** | **tiny-voca (M2)** | **4.5 M** | **0.733** | **4.34** |
|
| 203 |
+
| 9 | phonon-40m-en | 51 M | 0.737 | 4.45 |
|
| 204 |
+
| 10 | supertonic-3 | 100 M | 0.743 | 4.45 |
|
| 205 |
+
| … | … | … | … | … |
|
| 206 |
+
| 12 | PiVo-1.4M | 1.4 M | 0.767 | 4.33 |
|
| 207 |
+
| **13** | **tiny-voca (F2)** | **4.5 M** | **0.773** | **4.43** |
|
| 208 |
+
| 14 | LFM3-Audio-2B | 2.05 B | 0.797 | 4.45 |
|
| 209 |
+
|
| 210 |
+
[Open TTS Leaderboard (16 Nov 2026 snapshot)](https://huggingface.co/spaces/hf-audio/open_tts_leaderboard?version=v1.6.3)
|
| 211 |
+
|
| 212 |
## Limitations
|
| 213 |
|
| 214 |
- English only, two fixed voices, no voice cloning.
|