NVIDIA text-to-speech models & pricing
1 model listed · appears on 1 of 3 boards
Publishes the open-weights Magpie multilingual text-to-speech models alongside its speech AI toolkits.
Models on the boards
ModelReleased$/1M charsBoardsComposite
Magpie-Multilingual 357M (Feb 2026)Open weightsFeb 2026—1/331.0
Latency from public benchmarks
Time to first audio, in milliseconds, as each benchmark publishes it. We don't measure latency ourselves, and the sources measure differently (harness, region, sample size), so each figure stays under its own source. Not part of the composite score.
None of the latency benchmarks we read lists a NVIDIA model yet.
Coval: Time to first audio, p50, us-east-1, read Oct 11, 2026 · Speko: Time to first audio, p50, US East, read Oct 11, 2026 · Humanness Index: Time to first audio, median, read Oct 11, 2026. How latency is sourced.