• 01

  Hook in the first 2 seconds or lose them

  Shorts retention is decided before second 3. Open on the punchline, the question, or the most counter-intuitive fact in the script. "Most people don't know that..." beats "Today we're going to talk about..." every time. Cut anything before the hook in the edit.

• 02

  Run TTS at 1.1–1.2× for short-form pacing

  Default playback sounds slow against fast cuts. Bump speed to 1.1× for explainers, 1.2× for high-energy hooks. Anything past 1.25× starts to sound chipmunked and viewers bail. Generate first, then nudge speed in the editor so you can A/B without re-rendering.

• 03

  Aggressive silence cutting

  Open the WAV in your editor, snap the playhead to every gap longer than ~250 ms, and ripple-delete. Even Kokoro inserts breath pauses that work in long-form but stall a 45-second clip. Tight gaps = tight retention curve.

• 04

  Build a 60-second time budget on paper first

  3 seconds hook + 40 seconds body + 15 seconds payoff/CTA = 58 seconds. At 150 wpm that is roughly 145 words, ~830 characters — well under the 5,000-character generation cap. Write to the time budget before you generate, not after.

• 05

  Avoid the over-circulated default voices

  CapCut's default and TikTok's built-in TTS voices are recognized within half a second by any active short-form viewer. The instant they hear it the brain registers "another faceless clip" and the swipe rate spikes. Pick from the Kokoro pool — Sky, Puck, Echo — and your hook gets a fair hearing.

• 06

  Sync captions to the audio, not the script

  Auto-caption the rendered WAV in CapCut or Premiere — do not paste your written script. TTS pacing differs from typing pacing, and out-of-sync captions kill perceived production quality faster than anything else.

In-app TTS evolves; check current platform terms before assuming commercial-use coverage on built-in voices.