Add a feeling where it belongs
In Text to Speech, place the cursor before a line and open Add a feeling. Choose a tag to insert it into the script. Tags are highlighted so you can distinguish directions from spoken words.
A tag requests a style; it does not guarantee an exact performance. The same direction can sound different on different voices. A short test is more useful than filling a whole script with tags.
[calm] Let us take this one step at a time. [excited] We made it to the final round!
Use the tags in the editor
The current menu offers these feelings and delivery cues. Insert the exact wording shown, including the square brackets.
| Type | Tags |
|---|---|
| Mood | [happy], [sad], [angry], [excited], [calm] |
| Attitude | [nervous], [confident], [surprised], [relaxed], [sarcastic] |
| Delivery and reactions | [whispering], [shouting], [laughing], [sighing], [crying loudly] |
Compare a plain take with a directed take
First, make the wording work on its own. Then add one cue at the moment the delivery should change. Keep the voice and Speed the same so you can hear what the cue changed.
The lights went out. For a moment, nobody moved. Then we heard someone laughing in the hallway.
Add one deliberate change
This is an example prompt, not a promised result. If the transition is too abrupt, remove a tag or generate the two moments as separate clips and assemble them in your editor.
[whispering] The lights went out. For a moment, nobody moved. [surprised] Then we heard someone laughing in the hallway.
Use words and settings together
- Pacing: punctuation and short sentences help shape a line. Use Speed for the overall pace, from 0.5× to 2×.
- Loudness: Volume changes the output level, from -5 dB to +5 dB. It is different from asking the voice to shout.
- Consistency: Loudness normalization helps manage output levels. Listen to the final audio in the context where it will play.
- Numbers and dates: Text normalization is available in Settings. Still check important names, dates and amounts by ear.
- Advanced: Temperature and Variation are available under Advanced. Leave them at their defaults while learning how a voice responds.
If a tag does not sound right
Avoid copying tag syntax from a different service. Parentheses, XML pause markup and instructions for older speech models are not interchangeable with the current feeling menu.
Tags count toward the 5,000-character Text to Speech limit. Check the displayed credit cost before each Generate.
- Check that the tag uses square brackets and the intended spelling.
- Remove competing directions from the same sentence.
- Make the words support the requested feeling.
- Try the same line without the tag to isolate the problem.
- If necessary, try another voice or split the script into shorter sections.
Frequently asked questions
Do I need a tag on every sentence?
No. Start with plain text and add a tag only where you want a clear change in delivery.
Does [shouting] just increase the volume?
No. It requests a vocal performance. The Volume setting adjusts the output level separately.
Can I force an exact emotion?
No. Tags guide the generation, and results vary by voice and take. Review the audio before using it.