Text to speech for YouTube, from script to export.

Write the script for listening

A script that reads well on a page can sound stiff when spoken. Read it aloud once. Where you run out of breath, add a full stop or split the sentence. Use contractions (“it’s”, “you’ll”) if that is how you talk, and keep one idea to a sentence.

Pick a voice that fits

Match the voice to the video: calm and warm for tutorials and explainers, upbeat for reviews and lists. Every voice in the voice library has a sample, so spend a few minutes listening. If you publish regularly, stick to one voice so viewers get used to it. To have a voice that is only yours, clone your own.

Generate in sections

A single generation can be up to 10,000 characters, about ten minutes of speech. For a long video, generate it section by section: if you change one paragraph later, you only regenerate that part. With unlimited text to speech, redoing a section costs nothing extra.

Tune the delivery

• Slow the speed a little for complex topics, and speed it up slightly for energetic content.
• Use the calmness control to make the delivery steadier or livelier.
• Spell out anything that is read wrongly, such as acronyms and unusual names.
• Generate a short test sentence first, then the whole section.

Add it to your video

Each generation downloads as a standard WAV file. Drop the files onto the timeline in your video editor, line them up with your visuals, add background music at a lower volume, and export. If you recorded some lines yourself, the free Voice Enhancer can remove noise and even out the volume so that everything matches.

Common mistakes to avoid

• Pasting a very long script in one go. Generate it in sections instead, so changing one paragraph is quick.
• Skipping the test sentence. Always hear a short sample before you generate the whole section.
• Leaving numbers, dates and acronyms as they are. Write them the way you want them spoken, for example “twenty twenty-six”.
• Using one flat delivery from start to finish. A slightly slower intro and a livelier middle keeps people listening.
• Forgetting to download. Your browser keeps only your last five generations, so save each file as soon as it is ready.

Be honest with your audience

Platforms have rules about AI-generated and synthetic content, and some ask you to label realistic synthetic audio. Check the current rules of the platform you publish on. Never use a cloned voice of a real person without their permission.