1. Punctuate for breath, not just grammar
Punctuation is how the AI learns to pace and pause. Add commas where a person would take a small breath, and use periods to create full stops. Breaking a long run-on sentence into two shorter ones instantly sounds more natural.
2. Start with a style preset
Instead of guessing at technical sliders, pick a preset that matches your content — Conversational for casual talk, Narration for storytelling, or Newsreader for clear delivery. Presets tune expressiveness and stability for the scenario so you get a natural result on the first try.
3. Match the voice to the content
A warm, relaxed voice suits an audiobook; a bright, energetic one suits social media. Preview a few voices and pick the one whose natural character already fits your script — you'll fight the settings far less.
4. Fix pronunciation with spelling
If a name, brand, or acronym comes out wrong, respell it phonetically (for example, write 'nih-KAY' instead of 'Nike' if needed) or add spaces and hyphens. Small spelling tweaks reliably fix mispronunciations without any special markup.
5. Don't over-crank expressiveness
Maximum style and emotion can sound theatrical or unstable. For most narration, a moderate setting reads as more genuine. Raise expressiveness only for lines that truly need it, and keep longer passages steadier.
6. Adjust speed to the context
Slightly slower pacing suits explainers and audiobooks; a touch faster suits upbeat promos. Small speed changes have a big effect on how relaxed or urgent a voice feels — nudge, don't slam.
7. Generate, listen, and iterate
Read the output out loud in your head first, then generate and listen. When a line sounds off, tweak that sentence's punctuation or spelling and regenerate just that part. A couple of quick passes gets you to natural.