In September 2025, 40 early-access users answered the same writing prompts twice, once by keyboard and once by voice, so we could compare the two methods side by side. Our speed claims are based on internal benchmarks across 40+ early-access users across varied writing tasks.
The result was not one universal speed figure. Voice came out ahead overall, yet the size of that lead changed with the assignment, and some situations gave the keyboard a real chance to keep up.
How We Ran the Comparison
We used a straightforward setup. Every participant handled four tasks: a short email reply (under 50 words), a longer email written from a prompt (around 200 words), meeting notes based on a described scenario, and a free-form reflection on a topic they selected. Each task was done twice, once with a keyboard and once with voice, with the order randomized. We timed each attempt from the appearance of the prompt until the participant marked the work complete.
The group represented different typing speeds and professional backgrounds, and we did not filter for typing ability. Some participants typed around 40 words per minute, while others typed over 80. In the voice condition, everyone was using Wispr Flow for the first time and received a short orientation beforehand.
Tasks Where Voice Pulled Ahead
Voice finished faster on both longer assignments, meeting notes and the extended email, showing a steady advantage across the group. That pattern did not apply to every person on every task, but it was clear across the cohort.
Meeting notes produced the most dependable gap. People speaking from a described scenario finished with a meaningful time advantage over those typing. Independent reviewers also found that the spoken versions covered the scenario more fully, retaining details that typed versions sometimes left out or shortened.
The longer email followed the same general pattern. Most participants dictated it in one forward pass, while keyboard versions included more pauses, deletions, and rewrites. Typing also led to more edits inside sentences and more changes to the overall structure, increasing completion time before the draft felt finished.
Where the Keyboard Stayed Competitive
For the short email reply, the group showed almost no reliable difference between voice and keyboard. On these brief responses, faster typists were often quicker with the keyboard. Starting voice input and adjusting to speaking a one-sentence answer reduced the benefit voice tends to show on longer work.
That gives the result an important boundary: voice is not faster for every task. On very short output, a skilled typist's per-character speed can erase or reverse the gap. With longer prose, voice moves back ahead because typing speed does not build quickly enough to offset the raw speaking rate.
Participants typing over 70 words per minute had a smaller voice lead across the task set. Voice was still usually faster, but by less. For participants below 55 words per minute, the voice advantage was more pronounced across the board.
A Notable Result: Editing After the Task
Editing time was not part of the original study design. Several participants raised it in their post-task comments, though, and we tracked it informally during the final sessions.
People who created their drafts by voice spent less time changing the substance afterward. Their first pass was closer to the intended message. Keyboard drafts contained more incomplete sentences, missing phrases, and structural issues that required later revision.
Our current explanation is that fast typing can break sentence construction into fragments that must be repaired later. Speaking encourages more complete sentence arcs because the speaker follows the sentence as a whole in real time. Flow's cleanup layer removes filler sounds, leaving prose that is closer to finished without requiring the same cleanup work.
We are not presenting this as a universal rule. It is an observation from 40 users across four task types over a short study period. Still, the pattern appeared often enough to report.
What the Comparison Left Out
There are meaningful limits to these results. Forty users in a short controlled session do not represent a longitudinal study of everyday work. The assignments were built to resemble knowledge-worker writing, but they were constructed scenarios rather than real email threads or actual meeting notes.
The participants were self-selected early adopters, so they were probably more willing to try voice input than a random population. They also completed the voice tasks after only a short orientation. People with more practice would likely show a larger voice advantage than we measured.
We did not test work that naturally favors a keyboard, including structured data entry, form completion, and short precise edits to existing text. The comparison focused on prose generation, the use case Wispr Flow was built for.
How to Apply the Findings
Use these results as a direction, not a rulebook. Instead of asking whether voice is always faster by one fixed percentage, ask where voice writing could save meaningful time in your own workflow.
Our strongest candidates are longer prose tasks built from memory and knowledge: email replies over 100 words, meeting notes, working summaries, and any assignment where the idea is already in your head and needs to reach a document.
For very short responses, structured work, and tasks that depend heavily on reference material, the keyboard remains compelling. For the space between those cases, trying voice is reasonable, and an honest experiment will tell you more than a universal claim.