AI voice studio
A guided voice studio with one decision per step, where the AI’s behavior is visible and controllable at every step.
Why it mattered
Narration was the bottleneck for training videos, announcements and e-learning. Every recording meant booking a speaker and a studio, and changing one sentence started the process again. The goal: let any content owner generate publishable narration inside the tools they already use, without a training course.
Who we designed for
The content owner
Writes the script and owns the deadline. Afraid of “robot voice” and of publishing a mispronounced name.
The trainer
Needs pauses in the right places and a way to slow a voice down for learners in a second language.
The platform admin
Cares about cost per generated minute, where files end up, and tenant policy, dark mode and accessibility.
Where it got stuck
Hundreds of voices with technical names, and no idea which to pick.
Names, places and specialized terms pronounced wrong, with no way to fix them.
Generated files went into email, then got lost.
Fear of wasting generation credits on something they could not preview.
The decisions
Four steps, one decision each
A wizard adds clicks but removes the blank-page problem. Each screen asks one question with one sensible default.
Dependent controls disable themselves
The interface never offers an option that cannot apply, and the rules are written on the design as developer notes.
Preview before you spend
“Play sample” generates a short clip first. Full generation is a separate, explicit action.
Pauses are visible objects
Color-coded chips in the text replace markup tags, so people see the rhythm at a glance.
Speech tuning in context
Speed, pitch and pronunciation live inside the script step, next to the words they affect.
Files land where work happens
Download or save straight to cloud storage. On mobile the design offers download only, and says so.