The obvious promise is speed. Customer evidence points to a sharper territory: long-form work without the rewrite.
The exact pipeline we'd run on this account: raw customer signal to a batch of testable creative, and the loop that keeps it going.
We coded 325 comments and reviews line by line, each tagged to the tension underneath it. A slice of the raw pile:
325 comments, coded across Reddit, Google Play, the App Store and Product Hunt. The brass lines mark one tension: post-dictation cleanup and accuracy.
A feature isn't a tension. Coded down, six knots kept surfacing.
Before choosing a message, we choose a person. The same comments map to several personas. Four matter for a first campaign; we'd build for the two that lead.
Professionals who write detailed emails, updates, notes and first drafts all day.
Typical use cases Long emails, detailed updates, work messages, planning, note-taking.
Minimize the time it takes to turn a detailed thought into a message that’s ready to use as-is.
“I can write out detailed responses to my team without having to call them.” Google Play
People who use ChatGPT, Cursor or other LLMs heavily and regularly create long, context-rich prompts.
Typical use cases Long prompts, coding sessions, brainstorming, explaining logic to AI.
Minimize the time it takes to get a full, context-rich prompt into an LLM without losing your meaning.
“Where it's most useful is structuring a very long prompt to any LLM.” Reddit
We scored six messaging territories for the primary persona, the long-form knowledge worker, and built the reel on the one that came out on top.
How we scored. Frequency = how often the idea appears in persona-aligned comments. Severity = inferred from the actual wording ("game-changing", "dangerous", "repair every few sentences", abandonment). Priority = frequency 60% / severity 40%.
Typing becomes the bottleneck when the message, update, note or draft gets long.
Speak the full thought at natural speed and get the long-form communication down faster than typing.
Pain / intensity evidence“responding to emails and Teams messages while walking or on the go”Reddit
If users must repair punctuation, wording or recognition errors, the time advantage disappears.
Turn natural speech into text close enough to use immediately, with minimal cleanup.
Pain / intensity evidence“it keeps up with normal speaking speed without garbling things”Product Hunt
Long replies and ideas get postponed until the user is back at a proper keyboard.
Respond, capture and draft while walking, driving or away from the desk.
Pain / intensity evidence“I can even use it while listening to music, i.e. driving”Google Play
Turning an idea into typed language can interrupt the thinking itself.
Speak first drafts, explanations and ideas without breaking momentum to type.
Pain / intensity evidence“Instead of stopping my flow to type everything manually, I simply speak my thoughts”Product Hunt
Once dictation enters real work, failed or incomplete long transcriptions can break the workflow.
Make long-form voice input dependable enough to become part of daily work.
Pain / intensity evidence“Very good when it works, but every single day I waste time troubleshooting”Google Play
Work spans email, chat, notes, browsers and editors; a single-purpose tool adds friction.
Use the same voice workflow across the apps where work already happens.
Pain / intensity evidence“It formats differently whether you're in Slack, an email, or a code editor”Reddit
The reel combines the top two territories. Voice wins the moment typing becomes the bottleneck on long emails, updates and drafts, but only if the output is clean enough to send without a cleanup tax. So we position Wispr Flow for long-form work, without the rewrite.
The first three seconds. The line that earns the rest of the ad. Same message, six different doors in.
The message they all frame: when a message gets long, speaking beats typing, but only if it comes out clean enough to send as-is.
“I run a research agency, and a surprising amount of my day is just getting long thoughts out of my head.”
“Short replies I'll happily type. It's the long ones that used to sit in my drafts for hours.”
“I've tried voice-to-text apps so many times, and I always ended up going back to my keyboard.”
“Voice-to-text isn't really saving you time if speaking faster just means more editing afterward.”
“I wrote the same long message two ways, just to see which one I'd actually have to fix.”
“The first time an app gave me a long message I could send without editing, I didn't trust it.”
The test set. The bank runs deeper; these are the first round.
Most of the scroll happens with the sound off. The first frame has to make the argument on its own.
Ours puts the cleanup on screen: a messy dictated message, then the cursor deleting filler, fixing punctuation, rewriting a line.
First three seconds. Open on a messy dictated email. The cursor rapidly deletes filler, fixes punctuation, and rewrites a sentence.
Why it works The pain is legible instantly, and it pairs with the cleanup-tax hooks.
Also considered
A coherence note. The cleanup frame is a Territory-2 visual, so it pairs with the cleanup-tax hooks. If a Territory-1 hook wins, we'd open on a long-form frame like "hands off the keyboard" instead, so the first frame and the first line pull together.
The finished asset, 46 seconds. The footnote to the thinking above, not a replacement for it.
A hook test only teaches you something if one thing changes. We hold the narrative, visual and CTA constant and swap only the opening, so any difference is the door, not the edit.
| Hook (door) | Hook rate | CTR | Trial conversion rate | CPA |
|---|---|---|---|---|
| Identity | 32% | 1.7% | 9% | $37 |
| Threshold | 37% | 2.1% | 13% | $26 |
| Confession | 40% | 2.5% | 13% | $25 |
| Insight — winner | 43% | 2.8% | 15% | $21 |
| Experiment | 35% | 1.9% | 11% | $30 |
| Curiosity | 44% | 2.3% | 8% | $38 |
How we read it. Getting people to stop is easy. Getting them to sign up is what matters. Curiosity had the most stops but the fewest sign-ups. Insight got the cheapest sign-ups, which proves the cleanup angle was right. Threshold came a close second, so we test more long-form hooks next.
A win isn't the finish line. It's the brief for the next test.