Updated
What is the Gemini 2.5 Pro TTS API?
The Gemini 2.5 Pro TTS API is live on UnificAlly as google/gemini-2.5-pro-tts, from $0.00035 per second of generated audio. Gemini 2.5 Pro TTS is the quality-first voice in Google's 2.5 generation. It reads text in any of 30 prebuilt voices, detects the language on its own across more than 70 of them, and returns 24kHz mono audio. Two speakers can share a single request, which covers most interview and dialogue formats without stitching files together.
It is the steady one. Where Gemini 3.1 Flash TTS performs a script with inline stage directions, Pro reads it evenly from start to finish, which is what long-form work actually needs.
Gemini 2.5 Pro TTS API specs
| Model ID | google/gemini-2.5-pro-tts |
| Endpoint | POST https://api.unifically.com/v1/tasks |
| Request types | Text-to-speech, two-speaker dialogue |
| Voices | 30 prebuilt voices, by Gemini name or OpenAI-style alias |
| Languages | 70+, detected automatically, or pinned with a BCP-47 code |
| Output formats | mp3 (default), wav, pcm, opus, aac, flac |
| Speed | 0.25 to 4.0 |
| Dialogue | Up to 2 distinct voices per request |
| Rate for generated audio | $0.00035/sec |
Billing is per second of generated audio, with no subscription. The Gemini API pricing guide covers the rates across the Gemini family.
How to use the Gemini 2.5 Pro TTS API
- Create an API key on the API keys page.
- Send a POST request to
/v1/taskswith the model ID and aninputobject holding the text and voice. - Read the
task_idfrom the response, then check the task untilstatusiscompletedand readoutput.audio_url.
curl -X POST https://api.unifically.com/v1/tasks -H "Content-Type: application/json" -H "Authorization: Bearer YOUR_API_KEY" -d '{
"model": "google/gemini-2.5-pro-tts",
"input": {
"text": "Chapter one. The harbour was quiet that morning, and the boats had not yet come in.",
"voice": "Charon",
"instructions": "Read as a calm audiobook narrator",
"response_format": "wav"
}
}'
The response returns the task ID straight away:
{
"code": 200,
"success": true,
"data": {
"task_id": "abc123def456",
"status": "processing"
}
}
curl https://api.unifically.com/v1/tasks/{task_id} -H "Authorization: Bearer YOUR_API_KEY"
The finished task carries the audio URL:
{
"code": 200,
"success": true,
"data": {
"task_id": "abc123def456",
"status": "completed",
"output": {
"audio_url": "https://files.unifically.com/audio/abc123.mp3"
}
}
}
Gemini 2.5 Pro TTS API parameters
Provide exactly one of input, text, or dialogue.
| Parameter | Type | Default | What it does |
|---|---|---|---|
input | string | none | Text to speak. The language is detected automatically |
text | string | none | Alias for input |
dialogue | array | none | Dialogue lines as [{"text", "voice", "speaker"?}, ...], spoken in order. speaker is optional |
voice | string | alloy | Voice name, either a Gemini voice or an OpenAI-style alias |
response_format | string | mp3 | Output container: mp3, wav, pcm, opus, aac, or flac |
speed | number | 1.0 | Playback speed multiplier, 0.25 to 4.0 |
instructions | string | none | Style directions the voice follows but does not read |
language | string | auto | BCP-47 code to pin the output language |
speakers | array | none | Maps speaker names in input to voices for two-speaker dialogue. When set, voice is ignored |
The Gemini 2.5 Pro TTS API documentation has the full reference, with the voice list, language codes, and both dialogue forms.
Gemini 2.5 Pro TTS key features
Consistent long-form delivery. Pacing and tone hold across a full section rather than drifting paragraph to paragraph, so a chapter rendered in one call sounds like one recording session.
30 prebuilt voices. The roster runs from firm and informative (Kore, Charon, Rasalgethi) through bright and easy-going (Zephyr, Callirrhoe, Umbriel) to warm and mature (Sulafat, Gacrux), each usable in any supported language.
Two-speaker dialogue. One request renders both sides of a conversation, with each line mapped to its own voice and natural turn-taking between them.
Style instructions. A direction like "read this as a calm documentary narrator" shapes the whole take and never appears in the spoken output.
Six output formats. MP3, WAV, raw PCM, Opus, AAC, and FLAC, so the audio drops straight into an editing timeline or a streaming pipeline without conversion.
What Gemini 2.5 Pro TTS is best for
Audiobooks and long reads
Even delivery across thousands of words is what separates a listenable chapter from a tiring one.
Course and training modules
One voice holds across a whole lesson, so the module sounds recorded rather than assembled.
Documentation narration
Technical text read clearly, with a style instruction keeping the register consistent.
Interview-format content
Two voices in one request covers host and guest without post-production stitching.
Localized versions
The same script in more than 70 languages, detected automatically or pinned per locale.
Accessibility narration
Steady pacing on dense material makes long passages easier to follow.
Gemini 2.5 Pro TTS use cases
Turn a manuscript into a narrated audiobook chapter by chapter, holding one voice across the whole book. Produce course modules where the narrator sounds identical in lesson one and lesson twenty, months apart. Publish an audio edition of your changelog or docs, so users can listen to a release summary on a commute. Render interview-style marketing content with a host voice and a guest voice in a single call, then drop the file straight into an editor without aligning two separate exports.
Gemini 2.5 Pro TTS limitations
No audio tags. Bracketed directions written for Gemini 3.1 Flash TTS get read aloud here, so scripts have to be cleaned before they move between the two. Dialogue stops at two distinct voices per request. The 32,000 token context window means long material still has to be chunked. Pro also costs twice what Gemini 2.5 Flash TTS does per second, which adds up quickly on short, high-volume snippets where the quality difference barely registers.
FAQs
People also ask about Gemini 2.5 Pro TTS
Gemini 2.5 Pro TTS is Google's higher-quality voice in the 2.5 generation. It turns text into 24kHz speech using any of 30 prebuilt voices, detects the language automatically, and can render a two-person conversation in one request. It is built for narration rather than for performance-heavy character work.
Accounts get keys from the API keys page; copy one and pass it in the Authorization: Bearer <key> header. Every Gemini 2.5 Pro TTS job goes to https://api.unifically.com/v1/tasks as google/gemini-2.5-pro-tts, so the first request is the text plus a voice like Charon, and the same key reads the finished task.
The request names google/gemini-2.5-pro-tts and puts the text (as text or input), a voice, and optional instructions such as a calm audiobook narrator style inside input; response_format picks mp3, wav, pcm, opus, aac or flac. It runs as a task, so poll /v1/tasks/{task_id} until status reads completed and take the audio from output.audio_url.
Pick Pro for long reads where consistency matters, such as audiobooks, course modules, and documentation narration. Flash costs half as much per second and is the better default for short, high-volume snippets like notifications and confirmations.
No. Bracketed tags like [whispering] are specific to Gemini 3.1 Flash TTS and would be read aloud here. Use the style instruction field instead, which steers the whole read without appearing in the spoken text.
Yes, up to two distinct voices per request. Send a dialogue array where each line carries its own voice, or write the script with speaker names and pass a matching speakers map. Requests with a third voice are rejected.
30 prebuilt voices covering firm, bright, warm, breathy, and gravelly characters, and more than 70 languages detected automatically from the text. A BCP-47 code like fr-FR or ko-KR pins the output language when you need a specific one.
MP3 by default, plus WAV, raw PCM, Opus, AAC, and FLAC. The model renders 24kHz 16-bit mono, so WAV and PCM preserve exactly what it produced and the rest are encoded from that.
Per second of delivered audio, at $0.00035/sec, verified from provider usage after each generation rather than estimated from your input text. You pay for the audio that came back.
A session carries a 32,000 token context window, enough for several thousand words at once. Split longer material into chapters or sections and join the resulting audio files.
Gemini 2.5 Pro TTS guides and comparisons
Review pricing, limits, and tested outputs before running this model.
