TTS benchmark now counts audible audio, driving Gradium from 430ms to 214ms latency
mattturck · x · 2026-09-11
Benchmark org Coval found most TTS latency benchmarks stop the clock when any audio packet arrives — even if it's silence. After switching to time the first hearable audio, Gradium's latency jumped from 172ms to 430ms with zero model or infra changes: their audio arrived fast, but opened with up to 655ms of dead air.
Gradium used that signal to retrain their default model and now hits 214ms, about 170ms faster. Coval's advice: when comparing TTS providers on latency, check what the number actually measures.
More from Models
- DeepSeek 4.1 flash reportedly uses large ngram embeddings, echoing Qwen4 architecture — ccerrato147 · 2026-09-11
- ValsAI launches RSI Index, first third-party benchmark measuring how close AI is to self-improvement — JenniferHli · 2026-09-11
- Assistant Benchmark goes live: 61 assistants scored across 15 real-use dimensions — Scobleizer · 2026-09-11
- Devin's New Model Verdict: Not a Benchmaxxer, a 'Killer Execution Model' at $20/Month — brandon_galang · 2026-09-11
- Business Insider Asked ChatGPT, Gemini, Claude and Grok How AI Could End Humanity — coinfanking · 2026-09-11
- Claims resurface that Moonshot's Kimi distilled from Claude raw CoTs — xuanalogue · 2026-09-11