Text to speech
Text to speech with 1,292 voices, including 93 gender-neutral ones.
Paste a script, assign a voice to each speaker, fix the words the model would get wrong, and export an MP3. Long inputs are supported: up to 120,000 characters and 120 speaker turns per run.
Voice catalog
What is actually in the library.
Counts taken from the public catalog. The catalog is weighted towards English and Chinese because that is where document-to-podcast use is concentrated.
| Language | Voices |
|---|---|
| English | 988 |
| Chinese | 189 |
| Indonesian | 101 |
| Japanese | 8 |
| Spanish | 2 |
| German | 2 |
| Russian | 1 |
| Dutch | 1 |
| Total | 1,292 |
By gender: 805 male, 380 female, 93 neutral, 14 unlabelled. Browse the full library.
Gender-neutral voices
93 neutral voices, which is unusual in this category.
A searched-for gap in most text to speech tools: one consistent voice that does not read as male or female.
93 neutral voices, not 2
Most catalogs offer a token "neutral" option. The library has 93 voices tagged Neutral across the languages it covers, so a single character can stay consistent across a whole series.
Useful for characters, not just politics
Gender-neutral narration suits corporate explainers, documentary voiceover, children’s content, and any script where the speaker should stay unspecified.
Preview before you commit
Every voice has a sample. Filter by gender, age, and language, then listen before assigning it to a long script.
What the studio handles
Built for scripts longer than a paragraph.
Long-form input
A single run accepts up to 120,000 characters — roughly a 90-minute read at a normal pace — before you need to split the source.
Multi-speaker scripts
Up to 120 speaker turns or sections per run. Write dialogue, narration with asides, or a two-host exchange in one pass.
Pronunciation rules
Add rules for names, acronyms, and numbers so the model reads them the way you intend instead of the most common reading.
Delivery controls
Direction tags such as [curious], [professional], or [summary] shape how a line is read, and speed, volume, and loudness can be adjusted before export.
How it works
Four steps to an audio file.
- 1
Paste or open a script
Open TTS Studio and paste text, or start from a generated podcast script if you came from the document workflow.
- 2
Pick voices per speaker
Assign one voice for narration or several for a dialogue. Filter the library by language, gender, and age.
- 3
Add direction and rules
Insert delivery tags, correct pronunciations for names and acronyms, and set the reading speed.
- 4
Generate and download
Synthesize the audio, review it in history, re-run a failed section without redoing the whole file, then download the MP3.
Frequently asked questions
Is the text to speech free?
You can browse and preview all 1,292 voices for free. Audio generation uses credits: a new account gets 50 trial credits valid for 30 days, about 8 minutes 20 seconds of standard audio. After that, the Standard plan is $12 per month for 500 credits, roughly 83 minutes.
Which languages does the voice library cover?
1,292 voices in total: 988 English, 189 Chinese, 101 Indonesian, 8 Japanese, 2 Spanish, 2 German, and one each in Russian and Dutch. The catalog is weighted towards English and Chinese, which is also where the podcast and document workflows are most used.
Does it have gender-neutral voices?
Yes — 93 of the 1,292 voices are tagged Neutral, alongside 805 male and 380 female voices. You can filter the library by gender and preview each one before using it. Neutral voices are the practical choice when a character or narrator should not read as male or female.
How much text can one generation handle?
Up to 120,000 characters and 120 speaker turns or sections per run. That is a full report or a short book chapter. Longer sources should be split into parts; keeping each part to one logical section also makes a failed segment cheaper to redo.
Can I fix names and acronyms that are read wrong?
Yes. Pronunciation rules let you define how a specific word should be spoken, which matters for product names, surnames, and initialisms. Add the rule, regenerate, and check the section where it appears rather than the whole file.
Can I make a two-speaker conversation instead of narration?
Yes. Assign a different voice to each speaker and write the turns in the script. A dialogue format is available during document-to-podcast generation too, if you would rather have the script written for you first.
Related guides
Browse the Voice Library
Filter 1,292 voices by language, gender, and age, and listen to samples.
AI Podcast Generator
Turn a document into an editable podcast script, then voice it.
Free AI Podcast Generator
How the 50 trial credits work across podcast and speech generation.
Pricing
Trial allowance, monthly credits, and the yearly option.
Paste a script and hear it in a voice you chose.
Preview any of the 1,292 voices free, then spend trial credits only on the audio you keep.