Text-to-Speech Generation Speed
Speakvora returns library phrases instantly and generates new text in seconds. One test measured 6.5 s for Spanish and 8.7 s for French on 2026-10-08.
Library Phrases: Instant
Phrases already in the Speakvora library return audio immediately. The English library contains 7,730 distinct phrases in the public catalog, so common requests for existing text are served without delay.
Live Generation: Seconds
New text not in the library is generated on demand. In one test on 2026-10-08, Spanish generation took 6.5 seconds and French took 8.7 seconds. These are single observations, not benchmarks. Generation time may vary based on text length, language, and system load.
How It Works
When you call POST /v1/speak with text and a voice, Speakvora checks the library first. If the exact phrase exists, the URL is returned immediately. If not, the text is generated in the background and queued for delivery.
You can also use POST /v1/prepare to queue text for generation without speaking it yet. This is useful for warming up audio before you need it. Prepare requests are not billed.
Voices and Languages
Live generation works for 7 Kokoro languages: English, Spanish, French, Portuguese, Hindi, Japanese, and Chinese. The Kokoro voices that support live generation are af_heart, ef_dora, em_alex, ff_siwis, pf_dora, pm_alex, hf_alpha, hm_omega, jf_alpha, jm_kumo, zf_001, and zm_009. Premium and Chatterbox voices are library-only and do not generate new text.
Text Limits
Each request accepts 1 to 500 characters of text. Longer content should be split into multiple requests.
Frequently asked questions
Are library phrases really instant?
Yes. Phrases already recorded in the library are served immediately. The English library contains 7,730 distinct phrases in the public catalog.
How long does it take to generate new text?
In one test on 2026-10-08, Spanish took 6.5 seconds and French took 8.7 seconds. This is a single observation per language, not a guarantee. Time depends on text length, language, and load.
Can I prepare text in advance?
Yes. POST /v1/prepare queues text for generation without speaking it. Prepare requests are not billed, and queued text is ready in a few seconds.
Which voices support live generation?
Kokoro voices support live generation: af_heart, ef_dora, em_alex, ff_siwis, pf_dora, pm_alex, hf_alpha, hm_omega, jf_alpha, jm_kumo, zf_001, and zm_009. Premium and Chatterbox voices are library-only.
What languages can be generated live?
English, Spanish, French, Portuguese, Hindi, Japanese, and Chinese. All 7 Kokoro languages support live text-to-speech.
Related: API documentation, pricing, developer guides and more answers.