Text to speech voiceover: prepare narration for a video or presentation
A text to speech voiceover starts with a script that works when heard aloud. Slide bullets, production notes, and a written report are useful references, but they rarely make a finished narration without editing. Prepare the spoken wording first, choose one catalog voice for the generation, and review the actual audio before fitting it to your video or presentation.
Write a text to speech voiceover script for listeners
Begin with what the audience should understand at each point in the visuals. Draft complete spoken sentences rather than reading every slide label. If a chart supplies the detail, the narration can explain what to look for and why it matters. Keep visual directions in a separate document so they are not accidentally spoken.
Remove placeholders, editor comments, and instructions such as “insert example here.” The AI voice generator narrates the submitted writing. It does not know which lines in a pasted production document were intended only for the editor. Read your script aloud once before submitting it to catch sentences that are difficult to follow in one hearing.
Use explicit references where the picture alone would otherwise carry the meaning. “Compare the two totals on the right” gives listeners a useful cue, while “as above” may make sense only on a page. Check numbers and factual claims in the source before generation; a well-spoken error is still an error.
Prepare input within the text limits
The text to speech tool accepts typed or pasted text and TXT, Markdown, or PDF input. The editor and text upload path are limited to 10,000 characters. Split a longer script into meaningful sections before submission and keep the sequence in your own project notes.
For a presentation, one section per topic or group of slides can make revision easier. Avoid cutting halfway through a sentence simply to reach a character target. Use clear section names such as introduction, example, and conclusion in your saved project files. Only include those labels in the submitted script if you want them spoken.
PDF input follows a separate extraction path. It needs an extractable text layer and is capped at 100,000 characters. Extraction runs locally, but the extracted text goes to an AI provider for speech generation. For precise voiceover wording, preparing a clean script yourself often gives you better editorial control than using a report's page layout unchanged.
Choose a catalog voice and test real narration
Select an available voice using the catalog information in the voice generator. Each generation uses one selected catalog voice. Choose a representative passage containing a person's name, a technical term, a number, and a sentence similar in length to the rest of the script.
The quick browser preview is a listening aid, not a guarantee of the exact generated AI sound. Generate the sample and judge that audio. Listen for pronunciation, awkward pauses, and whether a listener can follow the explanation without reading the script. Keep the selected voice consistent across related sections when that suits the project.
If an abbreviation is unclear aloud, consider spelling out its meaning in the narration script. If a name sounds wrong, revise cautiously and verify that your adjustment preserves meaning. Do not assume punctuation supplies precise controls over timing or expression. Listen to each revised result rather than inferring the outcome from how the text looks.
Calculate credits before generating revisions
Speech generation requires sign-in and sufficient paid credits. The rate is 1 credit per 100 characters, rounded up. A 2,450-character script uses 25 credits; a script of exactly 2,400 characters uses 24. A 350-character test uses 4 credits. These examples show why a short pronunciation check can be useful before submitting a full project.
Count characters rather than words, and check the submitted text. Headings, repeated instructions, and extra material can make the narration longer than intended. Treat a new generation as another processing request when budgeting revisions. See pricing for paid access instead of assuming a test generation is included without charge.
Text is sent to a third-party AI provider for generation, and results are stored in account job history. The site does not store source content beyond processing. Use writing you have permission to process, particularly when a presentation contains confidential information. Generating speech is different from the separate local audio conversion workflow.
Download parts and assemble them with the visuals
Results provide playback and WAV downloads. Longer results can arrive in several parts. Save each part in order and compare its first and last sentence with the matching script section. Use filenames that include a sequence number and revision so you do not accidentally assemble a test take into the final presentation.
Each WAV part has an MP3 conversion option that runs in your browser. WAV can be useful for an editing step, while MP3 provides a listening copy. Check which format your video editor or presentation application expects. The speech results workflow supplies audio; arrange that audio with your visuals in the destination application.
Watch the complete assembled presentation. Check that a spoken reference arrives while the relevant diagram is visible, and that transitions do not cut off endings. If timing does not fit, revise the script or the visual sequence in your editor. Do not assume a fixed speaking duration from word count alone.
Keep the final script, reviewed audio parts, and assembled project together. For a smaller editing task, proofreading by listening offers a useful review method. If you are choosing delivery formats, the MP3 and WAV guide explains why a listening copy and an editing source serve different purposes.
Frequently asked questions
Can I paste a complete presentation document?
Prepare only the wording you want narrated. Remove production notes and check the 10,000-character editor limit. Divide longer scripts into meaningful sections before generation.
Does the browser voice preview match the generated output exactly?
No. Listen to actual generated audio before choosing it for the project. A representative sample helps you check pronunciation and delivery with the words you plan to use.
Can I download MP3 narration?
Yes. Results provide WAV parts, and each part can be converted to MP3 in the browser using its download option. Save and review all parts in sequence.
How are narration credits calculated?
Speech generation uses 1 credit per 100 submitted characters, rounded up. Sign-in and sufficient paid credits are required, including for short generated test passages.