Gradium's tiny on-device Phonon TTS tops Hugging Face's Open TTS Leaderboard
mattturck · x · 2026-11-07
Phonon, a small on-device TTS model from Gradium, ranks first on the Hugging Face Open TTS Leaderboard while competing against models up to 100× its size.
Per Gradium's reported results: first place on intelligibility in German and Spanish, top-three intelligibility in every evaluated language, and leading CPU streaming results on time-to-first-audio.
Hugging Face's Lysandre notes this shows why TTS models should be compared across dimensions — intelligibility (word/character error rates), generation throughput, and time-to-first-audio — rather than a single overall ranking.
More from Multimodal
- Higgsfield Miracle vectorizes any video into layers for real-time pixel-level editing — azed_ai · 2026-11-07
- NVIDIA's Lyra 2.0 turns a single image into explorable, consistent 3D worlds — NVIDIAAI · 2026-11-07
- StepFun's StepX-OPD merges capability learning and few-step distillation for diffusion models — stepfun-ai · 2026-11-07
- ByteDance's Seedance 2.5 video model lands in Adobe Firefly Boards — rufusd · 2026-11-07
- ASMR video made in just 7 prompts with ElevenLabs SFX v3 — austin_malerba · 2026-11-07
- LTX2.5 test footage: 'The Inquisitors' dystopia series demo — No-Judge-2848 · 2026-11-07