Write or import your script, choose a voice and select Generate speech. Listen to your take, refine the delivery and download the version you like.
How much text fits in a take?
You can use up to 5,000 characters in one script, including all dialogue turns. Split longer documents into sections, then download your takes together.
Can I create a conversation?
Yes. Open Multi-speaker, choose a voice for each turn and set the gap between lines. Your dialogue is combined into one audio file.
Where are my scripts and recordings stored?
Your script is sent to Microsoft’s online speech service for synthesis. The server caches audio for 48 hours, then removes expired files during its next cleanup. Finished takes are also saved in this browser. Clearing browser data removes these local copies. Members can save selected recordings to cloud and save projects to their account; those copies stay until deleted from the member workspace or the account is deleted.
Why might a voice be temporarily unavailable?
Voice availability depends on the upstream speech service. If a request fails, wait a little and try again. AI Text to Speech does not promise unlimited use or uninterrupted availability.
DOCUMENTS, CHAPTERS & LONGER IDEAS
Give a document a listening edition.
Bring a PDF or Word document, check its reading order and divide longer writing into natural sections. Keep each part within the studio limit and download a complete collection when you finish.
Search live voices by name, language or region. Choose one to start creating.
—Available voices
—Speech languages
—Regional choices
Browse voices by language
About the voice catalog
Explore the available voices and choose the delivery that suits your project. A regional label helps identify a voice’s language and locale; it does not promise a specific personal accent. Voices labeled Multi support multilingual reading. Preview a short sample and check your own names and terminology before generating a longer recording. Avatars are AI-generated fictional characters, not photographs of voice performers.
This website is independent of Microsoft. The catalog can change, and request limits and source availability apply.
Stored on this device, in this browser. Download the ones you want to keep.
A FEW POINTERS FOR A GREAT RECORDING
Write it. Make it sound like you.
Write for a listener
Keep each sentence focused on one idea. Read the script out loud once before generating it. Spell out unusual abbreviations and write numbers the way you want them spoken.
Use [[p700]] for a 0.7-second pause. Each pause can be up to 10 seconds, with 60 seconds of pauses in total and a maximum of 50 pause markers.
Build a conversation
Choose Multi-speaker and give each turn a speaker name and voice. Add lines in the order they should be heard. Set a short gap between turns for a natural back-and-forth.
Every turn counts toward the same script character limit.
From document to download
Upload a PDF with selectable text or a DOCX file up to 15 MB. Image-only PDFs need OCR first. Long documents should be divided into shorter sections.
MP3 is easy to share. WAV works well for editing. Add your own MP3 or WAV soundtrack from a finished take; the music is mixed at 15% volume.
Your data and the voice service
Scripts are processed by Microsoft’s online speech service to create your recording. The server caches audio for 48 hours and clears the job’s script when processing ends. Cleanup runs every five minutes.
Takes are saved locally in your browser. Download the recordings you want to keep. Your saved takes remain in this browser until you remove them or clear site storage.
AI TEXT TO SPEECH / MEMBER WORKSPACE
Your creative workspace.
A home for your scripts, signature voices and finished takes.
Understand where scripts go and what local history does and does not preserve.
The script leaves the browser
AI Text to Speech sends text to Microsoft’s online speech service to generate audio. The queue holds a script while it is being processed and clears the application payload when the job ends. Avoid entering material that should not be shared with an outside service.
There are two audio copies
A server-side cache lets the app reuse a recent result. A successful take is also stored in the current browser. The browser copy can be played after the server copy expires, but it is not synchronized automatically. Members can choose Save to cloud on a take to access an account copy on another device.
Availability can change
The voice list comes from the source and can change. An online service can reject or fail a request. Wait before trying again, or use another available voice; there is no promise of uninterrupted or unlimited access.
Keep your own important files
Download recordings you want to keep. Clearing site storage removes the browser’s local history, drafts and preferences. Use the privacy page to understand the current storage behavior and the contact form to report a problem.
Make a small test first
Choose a voice, generate a short take and listen before building a longer project. Scripts are processed by an online speech service; avoid material you should not share with an outside provider.