Feature · Cloud text to speech
Cloud text to speech without running GPUs.
PyAI Speak is a hosted speech API: send text with a scoped key and pay $0.04/min of audio.
What PyAI gives you for cloud text to speech
- Cloud text to speech on PyAI Speak runs synthesis on managed servers and bills $0.04/min of delivered audio, so nobody on your side rents GPUs or patches inference boxes.
- One request returns wav or pcm for apps, kiosks, cars and smart devices, or G.711 when the audio has to play on an 8 kHz phone line.
- Cloning a clean clip is free, and Voice Design builds a voice from a written description, both returning an ordinary voice id.
- Self hosting an open source model still wins when audio must never leave your own network, since PyAI Speak runs only as a hosted API.
What happens when PyAI renders an app prompt?
Choose one of the three cases below and press play. Each line is an illustrative script rather than a log.
Pick a moment, then press play
How does PyAI cloud TTS handle a request?
A request to the hosted service follows one fixed path from access check to invoice line.
Access check at the Edge
The PyAI Edge confirms the key, its speech permission, credit balance and rate limit in a single check.
Input validated
Speak checks the model alias, input length, emotion, format and sample rate, then maps the voice id to a voice and language.
Audio produced on PyAI hardware
Synthesis, loudness adjustment and conversion to wav, pcm or G.711 all happen on managed servers.
Usage settled
Billing closes on delivered seconds, reported back in the response headers.
Balance hits zero
Without auto recharge, a live key gets an out-of-credit error (402). Test keys never see 402. A small daily cap applies instead.
A leaked key is revoked
Revocation reaches the PyAI Edge in about a minute, after which the key is refused.
Who uses a managed cloud TTS API?
Teams that want speech output without staffing a speech model team, from app teams voicing reminders to publishers adding read aloud to every article.
IVR menus at 8 kHz
Admins of phone systems such as 3CX often ask which TTS suits IVR. Speak returns G.711 prompts for balances or wait times that play on 8 kHz lines unchanged.
Spoken reminders
An appointment app voices reminder messages with no model on its servers, paying only for the seconds rendered.
Read aloud with little code
A common ask is putting good TTS voices on a site with little code. The Vercel integration gives a Next.js site a Speak key with no secret copied by hand.
Narration in a cloned voice
Make one free clone, then render new scripts from the Console whenever a video needs a voiceover.
One codebase from sandbox to launch
A team building a speaker or kiosk tests spoken replies on sandbox keys at no cost. Swapping in a live key is the only change at launch.
Prompts and agent in one voice
A recorded app prompt and a live Omni agent can share one voice id, so users hear a single consistent speaker.
Cloud text to speech questions people ask
Is cloud TTS cheaper than self hosting a model?
Usually yes at low or uneven volume, because cloud TTS bills only for audio produced while a self hosted model pays for GPUs around the clock. PyAI Speak charges $0.04/min with no platform fee. At very high, steady volume self hosting can cost less, but your team then owns uptime.
Can I self host PyAI Speak?
No. PyAI Speak is offered only as a hosted API, with no on premises or self hosted edition. Organizations that require every audio file to stay inside their own network should run an open source model themselves and add their own audio formats, scaling and monitoring.
Can PyAI cloud TTS make IVR and phone prompts?
Yes. PyAI Speak returns G.711 audio that plays on 8 kHz phone lines, so IVR menus and hold messages need no conversion step. PyAI tests recorded 100% format and sample rate correctness. Prompts cost $0.04/min, and the same voice ids also run inside PyAI Omni voice agents.
How do I enable a cloud TTS API on PyAI?
Open a PyAI account, go to the Console and create a sandbox key with speech permission. Send text, a voice id and a format of WAV, PCM or G.711 to the PyAI Speak API. Sandbox keys never bill, and new accounts receive $50 of credit with no credit card.
PyAI is voice infrastructure. It hands you the audio and transcript, and does not replace your own models or review tools.