Build More Immersive Experiences: Text-to-Speech API Full Release

[hydr0h0mie]

[Update] August 11, 2026

@hydr0h0mie

Hey Creators! We’re happy to announce that the text-to-speech API now supports 10 languages in addition to English: Spanish, German, Italian, French, Chinese (Mandarin), Hindi, Japanese, Arabic, Korean, and Portuguese. You can listen to a sample of these voices in our documentation and learn more about how to use them in this post.

---

Hi Creators,

We’re excited to announce the full release of our Text-to-Speech API! We released the beta for this API at the end of June, and we’re now releasing an updated version with higher rate limits and observability dashboards.

This API enables you to instantly create narration and NPC dialogue for step-by-step tutorials or game announcements, allowing you to launch voice-enabled experiences rapidly and without traditional production costs or complex audio production workflows. You can choose from 10 predefined voices to personalize your experience, with more voices coming soon.

Example of a robot providing a player with clues

Using the Text-to-Speech API

Learn more about how to use the Text-to-Speech API in our initial DevForum post, or check out our guide docs and API docs (AudioTextToSpeech for real-time audio generation, and GenerateSpeechAsset to generate audio assets you can reuse) on Creator Hub.

What’s new?

New Rate Limits

We’ve heard your feedback that our initial rate limits were too restrictive. We’ve significantly increased this limit and switched from requests per second to requests per minute. Here’s the new formula:

max requests per minute per experience = 1 + 6 * number of concurrent users

This means that an experience with 10,000 concurrent users would be able to have 60,001 max requests per minute. This provides substantially more capacity than before. We hope this better supports your use cases and makes the feature easier to integrate in your experiences.

Observability Dashboards

We’ve also added Text-to-Speech observability dashboards on Creator Hub so that you can monitor API usage, troubleshoot issues faster, and make informed decisions about scaling. You can find this by opening your experience on Creator Hub and going to Monitoring > Text-to-Speech in the navigation menu on the left.

The first dashboard will show you how many requests your experience is sending, broken down by API type (real-time vs. asset generation). The top line shows the max number of requests you can send based on the number of concurrent users in your experience.

The second dashboard will show you the response and error rates of Text-to-Speech API calls, broken down by response status code.

You can adjust the date range and time intervals for the data at the top of the page.

What’s Next

Some upcoming additions to Text-to-Speech (TTS):

• Integration with Roblox Extended Services, so that you can pay for more usage if the free limit isn’t enough for your experience
• Support for additional languages to make TTS accessible to our global community
• Custom voice creation, so that you can create unique TTS voices that best fit the environment of your experience

This release is part of our broader vision for natural language-powered experiences. We’re also releasing the Speech-to-Text API beta by the end of year. When you combine Text-to-Speech, Speech-to-Text, and Text Generation, you’ll unlock completely new gameplay possibilities where players can have natural conversations with any NPC.

Please let us know if you have any questions, feedback, or would like to sign up for Natural Language office hours here.

Thank you!

FAQs

1. Is there any cost, request rate or asset quota limitations?

The API baseline utilization will be free. However, we have rate limits in place to optimize for game dialogue that also prevents system abuse as follows:

• Character Limits: A maximum of 300 characters per request is set for both versions of the API
• Request Rate Limits: We have enabled dynamic scaling based on your experience’s concurrent users using this formula: max requests per minute per experience = 1 + 6 * number of concurrent users
• Asset Generation Limits: Asset generation requests count toward your existing audio upload quota. This only applies to the GenerateSpeechAsset API

These limits help ensure system stability and fair usage across all users. Please note that longer content can be split across multiple requests.

2. What happens if my text input violates Community Standards?

The API will return an error and no audio will be generated. All text and audio output is filtered before processing to ensure compliance with Roblox Community Standards. In rare instances where violative content may be generated, the developer responsibility varies by API type as follows:

• AudioTextToSpeech API (real-time): You will not be held responsible for potentially violative audio outputs unless you intentionally program the system to generate violating content.

• GenerateSpeechAsset API (saved assets): Since generated audio assets are saved to your inventory, any moderation actions will be taken against your account as the asset owner. We strongly recommend against allowing players to directly input text for audio asset generation to avoid potential moderation issues.

3. Would including Text-to-Speech in my experience impact its Content Maturity rating due to the new AI Interaction policy?

No. Your experience’s Content Maturity rating will not be impacted solely because you include an AI component, including Text-to-Speech. Only an experience where users can engage in unlimited conversations with an AI model will receive a “Restricted” rating to help users and parents make informed decisions about appropriate content.

4. How many predefined voices are available for use?

We currently have 10 voices that you can use for this API. They each map to an numerical ID that you can input in the VoiceId field (note that this field accepts string values, so inputs should be formatted as “1”.

Voice ID	Description
1	British male
2	British female
3	US male #1
4	US female #1
5	US male #2
6	US female #2
7	Australian male
8	Australian female
9	Retro voice #1
10	Retro voice #2

[riftkuro]

finally out of beta! love to see it

[Mateufil (Mate)]

Thanks for fully releasing this API and increasing the rate limits!

[metatablecatmaid (metatablecat)]

In the closed beta, I mentioned that the TTS struggled with British place names (even with the British voices), which my test place used a lot of, for example, it’d pronounce Wolverhampton as “wolvert hammon” instead of how it should be pronounced, what’s been done to improve the pronounciations of these?

[TheDislikedGamer (Gamer)]

(The speech bubble above my head was meant to say “Hello, Mr. Robot.”)

[vfx_1]

Are the issues which make the entire Text-to-Speech API barely useable going to be addressed?AudioTextToSpeech:LoadAsync() yields forever when called server-side on a production server.

[7eoeb (beoe7)]

@hydr0h0mie

The API will return an error and no audio will be generated. All text and audio output is filtered before processing to ensure compliance with Roblox Community Standards. In rare instances where violative content may be generated, the developer responsibility varies by API type as follows:

So we don’t have to filter the text beforehand?

[cinthooran (Magnum)]

honestly really cool. i could see many games using this.
idk how voice snythesisers work but will there be a way to add our own custom voices in there

[pr_mary (primary)]

Are there any plans to add more voices in the near future? Currently, there are a decent amount of voices to use, but it may be restrictive for ambitious story-based games.

[epic_tank]

If im correct, this can be called client sided causing exploiters to eat up the rate limit. What can we do to prevent this?

[hydr0h0mie]

We’ve definitely taken note of this feedback - uncommon words and proper nouns are a known limitation of the API, just because the text-to-speech model isn’t familiar with these terms.

We’re continuously training and trying to improve the model behind the scenes. I can’t give a timeline since this requires some research and experimentation from us, but hopefully these types of issues will gradually resolve themselves over time.

[hydr0h0mie]

Sorry for the confusion - text-to-speech is meant to be a client-only API, so audio is always output on the client side. We’ll try to make this more clear in the documentation and/or make sure the API produces an appropriate error when it’s called server-side.

[hydr0h0mie]

Correct - the text filtering is done for you by default when you call the API.

[ldkWhatUserToPickk (Someone)]

THANK YOUUU!! I’ve been waiting for the rate limit increase since so much time

[Yal_f (yal)]

Will you ever consider adding more voices, or perhaps even custom voices? I feel the variety of current voices are too limited; multiple characters having the exact same voice can be pretty odd.

Edit: Saw that custom voices are planned, which is nice! More pre-defined voices can definitely save a lot of time though.

[ldkWhatUserToPickk (Someone)]

You can make a lot of voices with pitch and other configuration, well anyway, i’m back in the game you don’t have any chance

[hydr0h0mie]

@pr_mary

Are there any plans to add more voices in the near future? Currently, there are a decent amount of voices to use, but it may be restrictive for ambitious story-based games.

@Yal_f

Will you ever consider adding more voices, or perhaps even custom voices? I feel the variety of current voices are too limited; multiple characters having the exact same voice can be pretty odd.

We plan to add some additional voices in the future, but there likely won’t be any significant expansion. One way to make unique voices is to customize existing voices with audio effects - we also plan to allow users to create their own voices sometime next year.

[Mirtiny]

Will passing filtered text always guarantee that the API accepts the text?

[GoobsterX (Diamond)]

This update is exciting! I hope to see the different things that can come from it in the future and implement it into my own features!

I ask though will there a plan to expand the voice library by adding more of the “robotic” sounding TTS voices in the future? It would be interesting to see these kinds used for things other than realistic human npc’s, like computers or announcement systems, or just for a more generic voice.

(Examples taken from the website ttsMP3 using Amazon Polly)