elevenlabs/tts
ElevenLabs text to speech. Reads text aloud in one of 21 built-in voices, across 29 languages on the default model and up to 74 on Eleven v3.
Model Inputs
The text to speak. Billed per character; the pool's free accounts hold 10,000 characters each
One of the 21 built-in ElevenLabs voices. All are English speakers, but they read other languages accurately with a multilingual model, in an English accent
Synthesis model. multilingual_v2 is the most stable for long text (29 languages), v3 is the most expressive and understands [audio tags] (74 languages, 5,000 characters), flash/turbo are the fastest
Audio container, as codec_samplerate_bitrate. Higher-fidelity formats (mp3_44100_192, pcm_44100, wav_44100) need a paid ElevenLabs tier and are not available here
How steady the delivery is. Lower is more emotional and more variable between generations, higher is more monotone
How closely the output sticks to the original voice
Style exaggeration. Anything above 0 costs extra latency
Boost similarity to the original speaker, at a small latency cost
Speaking rate. 1.0 is normal; the API accepts 0.7 to 1.2 and refuses anything outside it
ISO 639-1 code forcing a language for synthesis and text normalization, e.g. 'ja'. Ignored by eleven_multilingual_v2, which detects the language itself
Whether numbers, dates and abbreviations are spelled out before being spoken. 'auto' lets the model decide
Best-effort deterministic sampling: the same seed, text and settings return the same audio. Determinism is not guaranteed
Result
Pricing
elevenlabs/
tts
Pricing for Synexa AI models works differently from other providers. Instead of being billed by time, you are billed by input and output, making pricing more predictable.
For example, generating 100 requests costs about $5.00.
Check out our docs for more information about how per-request pricing works on Synexa.
| Provider | Price ($) | Saving (%) |
|---|---|---|
| Synexa | $0.05 | - |
| fal | $0.05 | same price |
Readme
ElevenLabs text to speech, running on our own pool of ElevenLabs accounts. Give it text and one of 21 built-in voices and it reads the text aloud, with per-character timings recovered from the upstream with-timestamps endpoint.
Capabilities:
- 29 languages on the default
eleven_multilingual_v2, up to 74 on theeleven_v3models - Seven synthesis models spanning the tradeoff:
eleven_multilingual_v2for stable long-form,eleven_v3for the most expressive read (it understands inline[audio tags]),eleven_flash_v2_5for the lowest latency - Delivery control through
stability,similarity_boost,style,use_speaker_boostandspeed - Thirteen output containers, from
mp3_22050_32up tomp3_44100_128, plus PCM, WAV, Opus and telephony u-law/a-law seedasks for best-effort determinism, so the same text and settings return the same audio
Usage notes:
- All 21 voices are English speakers. They read other languages accurately with a multilingual model, in an English accent — there is no native-accent voice here
textis capped at 5,000 characters for the twoeleven_v3models and 10,000 for the rest; longer text is rejected before the request is madespeedis accepted between 0.7 and 1.2 onlylanguage_codeforces a language on the flash and turbo models;eleven_multilingual_v2detects the language itself and ignores it- High-fidelity formats (
mp3_44100_192,pcm_44100,wav_44100) are not offered — they require a paid ElevenLabs tier