Paste every prompt you have. Get every prompt back as a named file, in the 8 kHz format your phone system actually accepts — in fifty-nine languages, in the same voice, for the price of a subscription rather than a studio day.
No card. No trial clock. The free allowance returns every month.
The recording itself is an afternoon. Everything around it is the cost.
A voice actor, a studio slot, a round of notes. Then the same again when legal changes one sentence.
The original reader moves on, rates change, and prompt 41 no longer matches prompt 40.
Each new language means finding another native reader. Most teams never get past the second one.
Send the name your phone system already uses and get that name back. The alternative — five hundred files called by a UUID — is an afternoon of renaming before anyone can load them.
Export at 8 kHz, mono, in G.711 — μ-law or linear PCM. That is what Asterisk, FreePBX, Avaya and Cisco expect. Everyone else hands you a 24 kHz MP3 and leaves the conversion to you.
Send main_menu, press_1, after_hours and get exactly those filenames back. No renaming five hundred files that arrived called by a UUID.
Paste the whole menu, or upload the sheet you already keep it in. One character count before you commit, one progress bar, one ZIP at the end.
The same menu in Turkish, German, Arabic and Japanese, read by a native-sounding voice. The thing nobody attempts with voice actors, because the cost scales with every language.
Voice, style and delivery belong to the batch, not the line. Prompt 300 sounds like prompt 1 — and so does the one you regenerate next month.
Submit a batch from your own tooling, get a webhook when it finishes. Your prompt copy can live where it already lives and reach the phone system without anyone opening a browser.
One prompt per line. Put the filename first, a tab, then the text — which is exactly what two columns of a spreadsheet produce when you copy them.
One voice for the whole set, so nothing drifts. Choose MP3, WAV, or 8 kHz G.711 for the phone system. The character count is on screen before you commit.
Every file named as you named it. A prompt that needs rewording next month is one line, not another studio booking.
These ten files were not recorded for this page. The script below was pasted into Bulk Generation and came back named exactly like this — one submission, one ZIP, about a minute of finished speech.
01_hook.mp3Most text to speech sounds like a machine reading a list.
02_promise.mp3ProsodyAI is built to sound like someone who means it.
03_engines.mp3Three engines on one account: our own Prosody model, OpenAI, and Google Gemini. Choose the right one for each project, with no extra API keys and no second subscription.
04_languages.mp3Fifty nine languages, and every built in voice on every plan, with no limit on how many you use.
05_free.mp3The free plan is not a demo. One hundred thousand characters every month, roughly two hours of finished speech, and no card required.
06_clone.mp3Record a short sample and every generation can speak in your own voice, on the free plan too.
07_studio.mp3Then go further. Lay a voice over video, turn photographs into a narrated slideshow, burn in subtitles, or generate a whole clip with Veo.
08_bulk.mp3Running a call centre? Paste five hundred lines and get five hundred named files back, in the eight kilohertz format your phone system actually accepts.
09_close.mp3ProsodyAI. Text that sounds like it means something.
10_cta.mp3Start free at prosodyai dot ai. Sign up now.
No editing pass, no retakes, no audio software. The filename, a tab, the line. Swap the text for your own menu and the output is your menu.
Pasted into the Lines box
01_hook Most text to speech sounds like a machine reading a list. 02_promise ProsodyAI is built to sound like someone who means it. 03_engines Three engines on one account: our own Prosody model, OpenAI, and Google Gemini. Choose the right one for each project, with no extra API keys and no second subscription. 04_languages Fifty nine languages, and every built in voice on every plan, with no limit on how many you use. 05_free The free plan is not a demo. One hundred thousand characters every month, roughly two hours of finished speech, and no card required. 06_clone Record a short sample and every generation can speak in your own voice, on the free plan too. 07_studio Then go further. Lay a voice over video, turn photographs into a narrated slideshow, burn in subtitles, or generate a whole clip with Veo. 08_bulk Running a call centre? Paste five hundred lines and get five hundred named files back, in the eight kilohertz format your phone system actually accepts. 09_close ProsodyAI. Text that sounds like it means something. 10_cta Start free at prosodyai dot ai. Sign up now.
Style instructions
Confident, warm and unhurried. A brand voice, not a newsreader. Land the last line with quiet certainty, not a hard sell.
Plain language, not markup. It is how you direct a reader, and the premium engines take it the same way — which is also why a menu can be warm on the greeting and brisk on the options.
| Format | Audio | Use it for |
|---|---|---|
| Phone system — μ-law | 8 kHz · mono · G.711 μ-law WAV | Asterisk, FreePBX, Avaya, Cisco |
| Phone system — PCM | 8 kHz · mono · 16-bit linear WAV | PBXs that want uncompressed |
| WAV | Full rate · lossless | Archive and re-editing |
| MP3 | Full rate · compressed | Web players, hold music beds |
Most providers return 24 kHz audio and leave the conversion to you. Resampling after the fact is where prompts pick up the thin, tinny quality callers notice — so the conversion happens before mastering, not after.
Yes. ProsodyAI generates IVR voice recordings from text, in a voice you choose, and the paid plans include a commercial licence for the audio. The same voice reads every prompt, so prompt 300 sounds like prompt 1.
Most PBXs expect 8 kHz mono G.711 — either mu-law or 16-bit linear PCM — which is what Asterisk, FreePBX, Avaya and Cisco accept. ProsodyAI exports that directly, alongside full-rate WAV and MP3, so nothing has to be resampled after the fact.
Paste the translated lines and pick a voice. There are 59 languages across the three engines, and every built-in voice is available on every plan, so a second or third language costs the characters and nothing else.
Up to 500 lines in one submission. Put the filename first, a tab, then the text — which is exactly what two columns of a spreadsheet produce — and the whole set comes back as one ZIP, each file named the way you named it.
You edit one line and regenerate it. There is no studio booking and no voice actor to rebook, and the regenerated prompt matches the rest of the system because the voice and delivery belong to the batch.
Yes. There is a REST API and webhooks, so a batch can be submitted from your own tooling and a notification arrives when it finishes. Your prompt copy can stay where it already lives.
The free plan carries one hundred thousand characters a month — enough for a full menu and its translations, without talking to anyone. When it is time to put it live, the paid plans add commercial rights, the phone formats and the API.