💡 Based on Fish Audio S2.1 Pro. The tag list comes from the official docs. Every number in the tables below I generated and measured myself. Test set: Korean lines · 2 cloned voices · 244 generations. Results can differ in other languages or voices.

🔗 Fish Audio: https://fish.audio/?fpr=hyo-eun60


Remember these 3 and you're halfway there

  1. One tag, at the very start of the line. Adding a second tag at the end cuts the emotion in half.
  2. Each emotion has a different winning tag. [disappointed] beats [sad], and for irritation no tag works at all.
  3. Speed matters more than the tag. Pick the perfect tag at the wrong speed and it still sounds flat.

A. Start on the web (5 min)

  1. Go to fish.audio → Sign up
  2. Open the Text To Speech tab
  3. Pick a voice in the Voice field (defaults or the 2,000,000+ voice library)
  4. To use your own voice, upload a sample in Voice Cloning (15 seconds is enough)
  5. Type your line with an emotion in square brackets in front
  6. Generate & play
[disappointed] I confessed today... and got rejected...

The Tags button only shows 22 tags. The official docs list 70. The ones not on the button still work if you type them yourself. [disappointed], the strongest one in my tests, is not on the button.

Use Pro Controls to change speed and volume. Speed changes the emotion a lot, so always touch it.


B. Automate with the API (Claude Code / Codex)