Directing tone with text cues
In its tutorial, ElevenLabs presents two ways to steer generated speech: audio tags placed inside a script and Voice Changer, which works from a recorded performance. The audio tags shown in the video are short bracketed cues that suggest how a line should be delivered.
The examples try nervous, whispered and laughing delivery. The cue does not rewrite the words; it is intended to influence the manner of delivery. The tutorial also shows an Auto Tag control that can suggest such markers from the surrounding script.
Using a recording as direction
With Voice Changer, the user first records a short line in their own voice. The tutorial demonstrates the system attempting to carry the recording’s rhythm and delivery into a newly selected voice. That offers a different form of direction when a creator wants something more specific than a written mood cue.
The demonstration is an instructional example, not a guarantee of identical results in every project. Clarity of the recording, the selected voice and the needs of the scene can all affect the outcome. Comparing short tests is therefore more useful than treating one setting as a universal answer.
Choosing the right kind of control
Audio tags can be useful for quick script-level adjustments. Voice Changer can add another layer when a particular cadence or performance emphasis matters. The shared lesson in the tutorial is that voice output is shaped not only by the voice selection, but also by the direction given to it.
The practical question is simple: can the feeling of the line be described clearly in a few words, or is a performed reference the clearer brief?
