Grok Text to Speech and Voice

Grok text to speech is the voice control in the Grok app and on grok.com, not a separate TTS product page. The pricing page puts voice mode on the free plan’s bullets. SuperGrok Plus later mentions higher usage across Voice, which tells you the feature exists on more than one tier and that the meter still applies.

How to start it

Sign in, open a chat, and use the voice or headphone control in the composer. Allow the microphone if you want to talk, not only to listen. For a pure read-aloud of text you paste, type the text and ask Grok to read it in voice mode, then press the voice control. Headphones matter in a shared room because the reply may include text you did not mean to broadcast. Update the official app if the control is missing. Do not install an unofficial APK that advertises “Grok TTS unlocked.”

What you can expect

Voice is built for a conversation: a question, a spoken answer, a follow-up. It is a poor fit for a 40-page narration, a brand voice you must match, or a video soundtrack with frame-accurate cues. Imagine video has its own audio pass, documented in the video guide, and that audio is generated with the clip rather than exported as a clean voice track from chat. If you need a wav file for an editor, voice mode is the wrong tool.

Spoken answers drop links and hedges. Ask for the same answer in text when you need a source. A confident voice is not a citation. For translations you will send to someone else, use written translation and keep the text. Listening to a translation once is not a review.

Limits and plans

Free voice is real and capped. The cap is the notice in the app, as with other limits. Plus is the tier that publicly says voice usage goes up along with chat, Imagine, and Build. You do not buy a voice pack as a separate SKU on the pricing page we checked. Companion voices for Ani, Rudi, and Valentine are character voices inside those modes. They are not a general TTS voice picker for your own script.

API speech, if you are building a product, belongs to the xAI API docs for the current audio models, not to a guess from the consumer button. Consumer voice and API audio can ship on different schedules. Read the docs before you promise a voice ID to a customer.

A short spoken workflow

Paste the text you want read, in a normal chat, and add one line: “Read this aloud, do not add an introduction.” Start voice mode. If it adds a preamble anyway, stop and repeat the line. The point of the test is whether you can get a clean read, not whether the voice sounds like a studio. For a conversation, skip the paste and talk. For a translation you will send, do not use this path. Use the written translation page.

Background noise and a locked microphone are the usual failures, not a missing plan. The free tier lists voice. Check the OS permission, then the in-app control, then the plan notice if a quota sentence appears. Plus raises voice usage along with other products. It is not a voice-ID store. You will not find a public list of celebrity voices, and you should not try to clone a real person’s voice with a prompt. That request sits outside a normal product note and outside what this guide will help you do.

Companion voices are characters. If you open Ani, Valentine, or Rudi, you get their performance, not a neutral narrator. Switch back to standard Grok before you ask for a readout of a document. Otherwise the document comes back in character, which is a worse readout, not a better one.

Developers who need a file should read the current audio section on docs.x.ai rather than screen-record the app. A screen recording is a messy stereo file with UI clicks. The API, when an audio model is listed, is the supported automation. It bills separately from the consumer voice button. Do not assume the $0 plan includes API audio.

Use voice for intake, text for anything you store. A spoken brainstorm while you walk is a fair use of the free tier. The moment a sentence becomes a commitment, a caption, or a quote, switch to text and edit. Captions need spellings voice will not show you. If you record the device speaker with another phone, you also record the room and you still do not get a clean track. That workaround is a sign you wanted a file the consumer button does not offer. Stop and check the API docs, or book a human narrator, instead of shipping the room tone.

Accessibility is a real reason to use the control. If you need spoken replies because reading the screen is hard, the same free-plan voice mode is the feature, and the quota notice should be in text as well so you can see why speech stopped. A stopped voice with no explanation is usually a permission or a mute switch. Check those before you assume the plan removed the feature overnight.