Regeneration made repeated voice
generation faster.
Users can regenerate a result, adjust accent and pitch, and continue iterating without returning to the voice catalogue.
1
repeated loop
When a result was not right, users had to return to the catalogue, find the same voice, select an accent and pitch, start another generation and wait for a new result.
Accent and pitch often require several attempts. Support feedback showed that repeating the entire flow was slow and frustrating, causing some users to stop before reaching the result they needed.
2
from the result
I designed Regeneration directly inside the conversion list. Users can start from an existing result, update accent and pitch, and generate another version without searching for the voice again.
For TTS (Text To Speech) conversions, each regeneration produces a fresh result. Every voice includes ten free regenerations, making experimentation easier to understand and access.
3
in context
The regenerated version is processed alongside the original conversion and returns to the same list. Users can immediately play the new result, compare versions and regenerate again if needed.
The flow keeps the user’s voice choice and previous results visible, turning a repeated catalogue journey into a continuous generation loop.
For users, Regeneration removed the repeated journey through the voice catalogue and made frequent accent and pitch adjustments faster.
For the business, the shorter flow led to more generations and higher token usage because fewer users stopped after an unsatisfactory result.
After launch, internal product analytics showed a 32.3% increase in audio generations and a 27.2% improvement in happy-path completion.
“Oleksandr brought not just deep expertise in design, but a sharp product mindset. He pushed for practical prioritization changes, insisted on grounding decisions in user data, and advocated strongly for talking to real users.”
Mykhailo Kharchenko
Product Lead at Respeecher · Direct manager