Voice input is fast, but raw transcripts are messy, filled with errors, and often take longer to clean than typing. We designed Smart Dictate, QuillBot’s voice-to-refined-text feature, to deliver polished, ready-to-use writing by removing fillers, correcting grammar, and preserving tone. This transformed dictation from a novelty into a reliable way to write on mobile.
What was my role?
Product Designer, Strategist, and Usability Evaluator
Who did I work with?
Shubhashish Sinha (PM) and Christian Konrad (PM)
What did I work on?
Product Design, Strategy, Visual Design, Interactive Prototypes, and Usability Tests
Typing on mobile is slow and effortful, pushing users toward voice input. But existing voice-to-text tools are limited:
In QuillBot’s own ecosystem, 25% of Android keyboard users were already using built-in voice input — but with no way to polish it. This was a missed opportunity to demonstrate QuillBot’s AI value in everyday writing.
Provide ready-to-use, refined text that reduces cleanup.
Improve clarity without stripping away the user’s natural style.
Make writing easier for users who rely on voice input.
We built Smart Dictate with a three-layered system:
Google Speech-to-Text API converts speech into raw text.
QuillBot AI cleans and refines by:
Needed price markup considerations for mobile to balance revenue without harming trust.
Adapts refinement to surface (chat vs. email vs. docs).
Designing Smart Dictate wasn’t just about adding a mic button — it required nuanced choices:
Balanced making the voice CTA easy to find without cluttering the keyboard. Tested placement in quick-access bar vs. assistant menu.
Added live incremental transcription + progress animations so users could see text forming, not just wait for results.
Instead of auto-inserting refined text, added a preview/edit screen so users could review and trust the system.Instead of auto-inserting refined text, added a preview/edit screen so users could review and trust the system.
Replaced a static tutorial with inline microcopy + a “first dictation demo” to give users an instant aha moment.
Created lightweight cues (casual vs. formal) so refined text matched the writing context without manual toggles.
As the designer, I facilitated alignment across these teams, ensuring UX, tech feasibility, and AI behavior worked together.
Optimized latency and integrated the transcription API
Built models to refine text while preserving meaning.
Validated across accents, noisy inputs, and edge cases (long dictations, incomplete utterances).
Defined scenarios like casual chat, professional email, and noisy environments.
Dictated text was polished enough to use instantly with minimal cleanup.
Increased daily usage of dictation, shifting it from novelty to core input method.
Positioned QuillBot as smarter than generic voice typing, strengthening brand value.
.png)
.png)
.png)
.png)
.png)
.png)
.png)
Smart Dictate is the foundation for a hands-free writing assistant.
Future versions could allow natural voice commands like:
“Paraphrase that...”
“Make it more formal...”
“Translate this...”
Designing Smart Dictate taught me that voice features succeed not by being faster, but by being smarter and more trustworthy. By focusing on refinement, transparency, and context, we made dictation a reliable tool for real writing — and pushed QuillBot closer to its vision of being everywhere users write.
Back