The fastest text to speech API

Speakvora Phonon is 53 ms to first audio, ranked 3 of 37, at $3 per million characters.

Fastest first by median time to first audio. 37 models; the top 12 and ours are shown. Full table: speed and price.

#ModelProviderFirst audio, ms95th percentile, msPer 1M chars
1vuiFluxions3968$9.00
2Polly Standard *Amazon4277$4.00
3Phonon *Speakvora5396$3.00
4DefaultGradium56104not listed
5TTS Flash 2Inworld AI56111$15.00
6PhantomZ 4.0Deepdub6085not listed
7Qwen3 TTS FastNari Labs6597$10.00
8Kokoro *Speakvora8095$3.00
9Simba 3.2SpeechifyAI93127$10.00
10Pocket voices *Speakvora98132$3.00
11Simba 3.0SpeechifyAI99131$10.00
12Qwen3 TTS 1.7bBaseten105133not listed
15Polly Neural *Amazon143158$16.00

* Added by us with Coval's test method. Other rows: Coval TTS leaderboard, read October 30, 2026. Speakvora streaming needs a paid account.

The fastest at a low price

Polly Standard (42 ms, $4.00), then Phonon (53 ms, $3.00), then Kokoro (80 ms, $3.00), then Pocket voices (98 ms, $3.00), then Airy TTS (218 ms, $3.67). Speakvora is $3 per million characters, the lowest list price in this table.

Compare Speakvora with each provider

Prices side by side: the cheapest text to speech API.

Questions

What is the fastest text to speech API?

On the Coval TTS leaderboard (read October 30, 2026) the fastest listed model at the median is vui from Fluxions at 39 ms to first audio. Among the 37 models in our table, Speakvora Phonon ranks 3: 53 ms median and 96 ms at the 95th percentile, at $3 per million characters.

What is the fastest text to speech API that costs $4 or less per million characters?

In our table, the models at $4 or less per million characters are Polly Standard (42 ms, $4.00), then Phonon (53 ms, $3.00), then Kokoro (80 ms, $3.00), then Pocket voices (98 ms, $3.00), then Airy TTS (218 ms, $3.67). Speakvora Phonon is 53 ms at $3. Amazon Polly Standard is faster by 11 ms but costs $1 more per million characters.

How is first audio speed measured?

As the time from the request to the first audible sound: the arrival of the first audio chunk plus any leading silence, using Coval's published method. We tested Amazon Polly and Speakvora ourselves with that method on October 30, 2026; all other rows are Coval's, with the price column they list.

Does faster mean better?

No. Speed is one factor. We did not benchmark voice quality, accuracy or uptime, and Speakvora has no latency guarantee. Listen to the samples and test with your own text.