Start with the words you want spoken
A good Grok Voice prompt is a clear spoken script. In Text to Speech, the words entered in Your words become the material for your voiceover. The goal is to give the selected voice something natural to read: a focused message, understandable sentences, and wording suited to the listener.
This is different from asking a writing assistant to create a paragraph. An instruction such as “write a cheerful introduction for my video” describes a writing task; it is not the finished introduction. Prepare the actual narration first, then paste that narration into the voice generator. Keep filming notes, music cues, and instructions to your editor in a separate document.
Writing brief
Write a short welcome for a beginner photography lesson.
Text to paste
Welcome to your first photography lesson. Today, we will explore how light changes the way a photograph feels.
The examples below are original starting points you can adapt. They are not audio previews or promises of a particular delivery. Choose a catalog voice, generate a short take, and judge the recording against your project. For login, settings, and downloads, follow the How to Use guide.
Clear subject+One key idea+A useful ending=A focused voiceover script
Three formulas for a clearer voiceover
Before choosing a voice, decide who is listening and what they should understand afterward. A tutorial needs clear actions. An introduction needs context. A product message needs a specific benefit. Trying to do all three in one short paragraph usually makes the writing harder to follow.
Subject → Detail → Takeaway
“Today, we are looking at [topic]. [Explain one concrete detail]. Here is what to remember: [takeaway].”
Use this for a lesson or explainer. Name the topic early, illustrate it once, and give the listener a point to remember.
Need → Benefit → Next step
“Need a simpler way to [task]? [Product] helps you [specific benefit]. Start by [one action].”
Use this for a product video. Replace broad claims with a benefit you can demonstrate, and finish with one clear invitation.
Action → Result → Next action
“Choose [control]. You will see [result]. Next, [action] to [purpose].”
Use this for an app walkthrough. Match the actual interface labels and describe what changes on screen before moving on.
Replace every bracketed placeholder. These are writing templates, not generator commands. Paste only the finished narration into Your words.
Read the draft aloud before generating. If you lose track of a sentence, divide it. If several words are difficult to say together, rewrite the phrase. These edits improve the source material without depending on an unsupported control or hidden command. Keep each paragraph focused enough that you can replace it independently later.
Six text-to-speech scripts to make your own
Adapt these examples to material you have permission to use, following the usage terms.
Choose the example closest to your project. Copy the script, replace the details, and paste it into Your words. The notes beneath each example explain what to keep and what to change.
Video narration
A lifestyle opening with room for the visuals.
StructureSetting → Two visual details → Closing thought
Some mornings begin before the rest of the world is ready. The kettle warms. A window opens. For a moment, there is nothing to rush toward. Take a breath, find your rhythm, and make today your own.
Why this works. Short sentences separate the images: kettle, window, quiet moment. The final sentence gives the sequence a destination. Replace the setting with details that actually appear in your footage, then check the generated recording against your edit. These punctuation choices help communicate structure; they do not specify exact pause durations.
Product introduction
A clear benefit followed by a single next action.
StructureProduct → Everyday benefit → Invitation
Meet the notebook that keeps your day in one place. Capture a quick idea, plan your next project, and leave room for the things you discover along the way. Open to a fresh page. Start with what matters today.
Why this works. The opening names the product, the middle explains its use, and the closing gives one invitation. Adapt the benefits to your actual product rather than adding unsupported promises. If you need a shorter advert, remove one benefit first. Reading the remaining words faster is not a substitute for a more focused message.
Course or tutorial
Instructions a learner can follow one step at a time.
StructureFirst action → Next action → Reassurance
Start by opening your project folder. Create a new document and give it a name you will recognize later. Next, write one sentence that describes your goal. You do not need the perfect wording yet. You need a clear place to begin.
Why this works. Each sentence asks for one action. Words such as start and next make the sequence explicit without relying on a numbered list being read correctly. Replace generic actions with the exact labels learners will see. Listen to the result while performing the task, and check that the explanation leaves a natural place to stop.
Story opening
A specific scene that leads into a question.
StructurePlace → Unusual detail → Something changes
At the edge of the village stood a house with no front door. Every evening, a light appeared in the upstairs window. Nobody knew who lit it. Then, one rainy Thursday, a letter arrived with the house's address and Mara's name.
Why this works. Concrete details make this opening easier to picture than a general instruction to sound mysterious. The scene establishes a location, a repeated event, and a change. Keep character names consistent throughout the script. For a longer story, generate and review coherent scenes so a revised paragraph does not require rebuilding every section.
Podcast introduction
Tell listeners what this episode will help them understand.
StructureShow name → Episode topic → Listener takeaway
Welcome to Small Questions, a show about the choices we make every day. Today, we are looking at why a simple plan can be so hard to start. We will explore three common obstacles and one small change you can try this week.
Why this works. The listener learns the show name, the topic, and what to expect. Replace the example title and promise with your own. Read the show name in a short test before generating an entire introduction. Keep music and sound-effect directions in your production notes, outside the text you paste into the voice generator.
App walkthrough
Connect each visible action to the result it produces.
StructureControl label → Action → Review the result
Choose Text to Speech at the top of the generator. Paste your script into Your words, then select a voice and language. When you are ready, choose Generate voice. Listen to the completed recording, check the pronunciation, and download the audio for your project.
Why this works. This example uses the actual Grok Voice control names so the narration can match a screen recording. For another app, replace every label with the words shown in its interface. Split the narration when the viewer moves to a different screen. Check the finished audio against the recording rather than assuming the text length determines its duration.
Edit for the ear: before and after
Compare a revised take with earlier results in My Creations to hear how your edits change the delivery.
Text that works on a page can feel crowded when spoken. A listener cannot scan backward as easily as a reader. Give each important idea its own space and replace formal wording with the words your audience would actually use.
Before
In order to successfully complete the initial setup process, users are required to navigate to the settings panel and subsequently select their preferred language prior to proceeding.
After
Open Settings and choose your language. Then continue with setup.
Replace delivery directions with spoken words
Before · a direction
[Sound excited] Tell people our new planner is available.
After · a finished script
Your next idea needs a place to begin. Our new planner is here. Open a fresh page and make room for what comes next.
The second version supplies the actual message. It does not force an emotional delivery; choose a voice and listen to the take. Keep bracketed production notes outside the narration field.
In the setup example, the revised version preserves the action while removing unnecessary framing. It also gives the listener a clear order. This does not mean every sentence must be extremely short. Vary sentence length when telling a story, but keep instructions and important facts easy to recognize.
When a take feels awkward, change one thing at a time. Shorten a sentence, clarify a name, or remove an interruption. Generate the revised passage with the same voice and settings so the comparison tells you something useful. Save both versions until you have chosen the one that fits.
Make names, numbers, and language unambiguous
Start with the words that carry the most risk of confusion: product names, initials, dates, amounts, and specialist terms. Put them in a short test passage. Review the generated audio before producing a full lesson or a long narration.
| Source text | Possible spoken wording | What to check |
|---|---|---|
| 03/04 | March fourth | Confirm which date you mean. |
| 12.5% | Twelve point five percent | Listen for the decimal and unit. |
| Dr. Lee | Doctor Lee | Expand the title where appropriate. |
| Read the FAQ | Read the frequently asked questions | Choose whether to expand the abbreviation. |
These are editing options, not guaranteed pronunciation commands. For an unusual name, test a clearer spelling in a copy of your script and listen again. Keep the correct written form in captions and documents where spelling matters. Avoid replacing a term throughout a large project before checking one example.
Choose Language to match your script, or use Auto detect. The setting does not translate your writing. If you need narration in another language, prepare and review that language's text first. For a passage containing several languages or unfamiliar names, test the transitions and listen for missing or unclear words.
Use a recording as the starting point
Check the recording privacy guidance before uploading audio containing personal information.
If your ideas are already spoken, use Speech to Text to create a draft. Upload your recording or provide an accessible public audio URL. Choose the recording language and, when useful, add relevant names or specialist terms to Keywords. These keywords provide context for transcription; they are not instructions for a new voiceover.
Review the transcript against the source. Spoken conversation often contains repetitions, unfinished sentences, or references that only make sense in the moment. Correct errors first, then edit for the new audience. Keep an accurate source transcript separately if you also need a record of what was actually said.
Use as script transfers eligible text into Text to Speech. Once there, remove material you do not want narrated, choose a voice, and review the settings. Generate the new recording as a separate task. The full audio transcription walkthrough explains the upload and result controls.
A final check before Generate voice
- The text contains only words intended for narration.
- Names, dates, units, and abbreviations have been reviewed.
- The voice and language fit the script you are testing.
- The output format suits the app where you will use the audio.
- A separate copy of the script is saved for later revisions.
- You have checked the displayed generation estimate and youravailable credits.
Start with a small passage before committing to a long project. Each generation is a task, including a revised take or a voice comparison. Review the current credit packs and usage information when planning a series of recordings. Download approved outputs and name them so you can match each file to its script.
Grok Voice prompt questions
What is a Grok Voice prompt?
In this generator, it is the spoken script you enter in Your words. The interface provides separate voice, language, and audio settings. Write the actual narration in the text field and keep directions for a writer or editor outside it.
Can I paste an instruction such as ‘sound excited’ into the script?
Treat that wording as text that could be spoken. The current form has no separate delivery-instruction field. Express your message through the words you choose, select a catalog voice, and listen to a short generated sample before committing to a longer take.
Does punctuation guarantee an exact pause or speaking speed?
No. Punctuation makes sentence structure clearer, but the resulting delivery needs to be checked by listening. If your project requires a precise gap or duration, arrange and trim the downloaded audio in your editing software.
How long can my text-to-speech script be?
The current input limit is 60,000 characters. A short test passage is usually more useful while checking a voice or difficult term. Keep longer work divided into meaningful sections and preserve a separate copy of your complete script.
Can I use a transcript as a voiceover script?
Yes. Review the Speech to Text result, then use Use as script when the transcript fits the input limit. Edit the transferred words and select a voice before generating. Creating the new audio is a separate task.
Start with one good paragraph.
Choose an example, make the words your own, and listen to the first take. A useful script is one your audience can understand.
Try your script in Grok Voice