# Phonon update: 1.00% WER on Seed-TTS, smaller than every model we beat

DevFeed: [Phonon update: 1.00% WER on Seed-TTS, smaller than every model we beat](<https://devfeed.tech/articles/phonon-update-1-00-wer-on-seed-tts-smaller-than-every-model-we-beat-81241.md>)

Original publisher: [Read original article](<https://gradium.ai/blog/phonon-update-may-2026>)

Author: Gradium

Published: 2026-05-26T00:00:00Z

Content type: news

Language: en

Sources: [Gradium](<https://devfeed.tech/sources/gradium.md>)

Topics: [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [speech-to-speech](<https://devfeed.tech/topics/speech-to-speech.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [100m-parameter-tts-model](<https://devfeed.tech/tags/100m-parameter-tts-model.md>), [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-on-device](<https://devfeed.tech/tags/ai-on-device.md>), [allows](<https://devfeed.tech/tags/allows.md>), [api](<https://devfeed.tech/tags/api.md>), [applications](<https://devfeed.tech/tags/applications.md>), [approach](<https://devfeed.tech/tags/approach.md>), [apps](<https://devfeed.tech/tags/apps.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [best-on-device-tts-2026](<https://devfeed.tech/tags/best-on-device-tts-2026.md>), [browser-tts-wer](<https://devfeed.tech/tags/browser-tts-wer.md>), [device](<https://devfeed.tech/tags/device.md>), [edge-tts-model](<https://devfeed.tech/tags/edge-tts-model.md>), [english](<https://devfeed.tech/tags/english.md>), [int8-quantization-tts](<https://devfeed.tech/tags/int8-quantization-tts.md>), [kanitts2](<https://devfeed.tech/tags/kanitts2.md>), [mobile-tts-model](<https://devfeed.tech/tags/mobile-tts-model.md>), [on-device](<https://devfeed.tech/tags/on-device.md>), [on-device-tts](<https://devfeed.tech/tags/on-device-tts.md>), [on-device-tts-wer-benchmark](<https://devfeed.tech/tags/on-device-tts-wer-benchmark.md>), [parameter](<https://devfeed.tech/tags/parameter.md>), [phonon](<https://devfeed.tech/tags/phonon.md>), [phonon-vs-kokoro](<https://devfeed.tech/tags/phonon-vs-kokoro.md>), [phonon-vs-kokoro-wer](<https://devfeed.tech/tags/phonon-vs-kokoro-wer.md>), [phonon-vs-magpie](<https://devfeed.tech/tags/phonon-vs-magpie.md>), [phonon-vs-magpie-wer](<https://devfeed.tech/tags/phonon-vs-magpie-wer.md>), [phonon-vs-neutts](<https://devfeed.tech/tags/phonon-vs-neutts.md>), [research](<https://devfeed.tech/tags/research.md>), [seed-tts-benchmark](<https://devfeed.tech/tags/seed-tts-benchmark.md>), [seed-tts-english-benchmark](<https://devfeed.tech/tags/seed-tts-english-benchmark.md>), [small-tts-model](<https://devfeed.tech/tags/small-tts-model.md>), [speaker-similarity](<https://devfeed.tech/tags/speaker-similarity.md>), [speech](<https://devfeed.tech/tags/speech.md>), [supertonic-2](<https://devfeed.tech/tags/supertonic-2.md>), [text-to-speech](<https://devfeed.tech/tags/text-to-speech.md>), [text-to-speech-model](<https://devfeed.tech/tags/text-to-speech-model.md>), [tts](<https://devfeed.tech/tags/tts.md>), [tts-edge-model-comparison](<https://devfeed.tech/tags/tts-edge-model-comparison.md>), [voice](<https://devfeed.tech/tags/voice.md>), [voice-cloning](<https://devfeed.tech/tags/voice-cloning.md>), [wavlm-large-speaker-similarity](<https://devfeed.tech/tags/wavlm-large-speaker-similarity.md>), [whisper-large-v3-tts-evaluation](<https://devfeed.tech/tags/whisper-large-v3-tts-evaluation.md>), [word-error-rate-tts](<https://devfeed.tech/tags/word-error-rate-tts.md>)

## AI overview

Gradium reports updates to Phonon, its 100M-parameter on-device text-to-speech model. On the Seed-TTS English benchmark, it reaches 1.00% word error rate and 59.51% speaker similarity with voice cloning; with a fixed voice, it reports 0.83% word error rate. The update removes the prior minimum input padding and adds int8 quantization support, which the article says improves inference speed without perceptible audio-quality degradation. Phonon is in private beta.

## Source excerpt

Phonon, our 100M-parameter on-device Text-To-Speech model, reaches 1.00% WER on the Seed-TTS English benchmark, outperforming NeuTTS Air, KaniTTS2, and NeuTTS Nano. With a fixed voice, it drops to 0.83% WER, ahead of Kokoro and Magpie.