As an Engineering Supervisor at Buffer, I write loads — Slack messages, proposals, code, documentation, and suggestions. And as AI instruments turned a part of my workflow, I discovered myself writing much more.
I’ve at all times been gradual at typing, and I feel higher by speaking than by writing. It by no means felt like an enormous limitation earlier than, however with AI instruments, I began feeling it an increasing number of. I would maintain prompts quick even once I knew I ought to add extra element, as a result of typing all of it out felt tedious.
Then I attempted ChatGPT’s voice function and one thing clicked. I may embrace all of the context and nuance that I would usually skip. I wished that have in every single place, not simply in a single app.
So during the last yr, I examined 4 AI-powered speech-to-text instruments in my day by day workflow. They differ loads in velocity, accuracy, and options. Learn on to study extra about how I used every device, how a lot they value, and the way I combine dictation into my workflow.
Superwhisper: Highly effective however wants tinkering
Free plan: quarter-hour of dictation per 30 days
Paid plan: $8.49 per 30 days
For over a yr, I used Superwhisper and was fairly pleased with it. It is a strong device with a variety of customization – you’ll be able to configure completely different AI fashions, customized prompts, and modes for various purposes, so it applies completely different presets in Slack versus a coding IDE, for instance.
The subscription was inexpensive at lower than $10 per 30 days, and for some time, it labored nicely sufficient. However I bumped into two points. The accuracy wasn’t nice for my accent (I am not a local English speaker), and I usually wanted to make heavy edits after dictating, particularly with technical phrases. It was additionally pretty gradual – I would dictate one thing after which anticipate it to course of, which interrupted my movement.
I noticed I did not desire a device I wanted to tinker with and customise closely. I wished one thing that simply labored higher out of the field. That is what prompted me to strive different choices.
VoiceInk: Low-commitment option to strive dictation
$25 one-time fee
I attempted VoiceInk subsequent, based mostly on a teammate’s suggestion. It is a one-time buy of $25, and it is open-source, which I like supporting.
The consumer expertise is intelligent – it integrates immediately into the macOS notch. It is also totally native, so all processing occurs in your system, which is great from a privateness standpoint. It has some customization choices and was noticeably sooner than Superwhisper.
For my wants, the accuracy nonetheless wasn’t fairly there with my accent and talking model, which is why I saved trying. However if you happen to’re new to dictation instruments and need to experiment with no subscription dedication, this can be a very strong alternative. The worth level makes it low-risk, and for infrequent use or you probably have completely different accuracy wants than I do, it may work very well. Many people at Buffer use VoiceInk.
Wispr Circulate: Quick, feature-rich and polished
Free plan: as much as 2,000 phrases per week
Paid plan: $15 per 30 days
Wispr Circulate is the device I am at present utilizing.
It is considerably dearer than the earlier choices, however the accuracy is noticeably higher. It catches technical phrases reliably and mechanically saves phrases I’ve corrected, so it will get higher over time.
The most important distinction is velocity. It processes your audio whilst you’re nonetheless talking, so the transcription seems nearly instantly after you end. This has been transformative for my workflow – I can keep in movement and context change a lot much less incessantly.
It additionally handles a number of languages easily, which many different instruments do not. I often want to modify between English and Ukrainian (my native language), and Wispr Circulate manages this with none points.
At $15/month, it is the priciest choice I examined, however the efficiency and reliability have been price it for day by day skilled use.
Willow Voice: Nice high quality, however fewer options
Free plan: as much as 2,000 phrases per week
Paid plan: $15 per 30 days
Willow Voice is equally priced to Wispr Circulate and has lots of the similar advantages – it is fairly quick and has fairly good accuracy. For my specific use circumstances, it was barely slower and barely much less dependable than Wispr Circulate, however not by an enormous margin.
It looks like they seem to be a bit earlier as an organization, and a few options I anticipated weren’t there but. I am maintaining a tally of it, however I settled on Wispr Circulate because it was barely higher in a lot of the points that mattered to me.
| Device | Worth | Platform | Pace | Accuracy |
|---|---|---|---|---|
| Superwhisper | $8/mo | MacOS + iOS | Medium * | Medium * |
| Voice Ink | $25 one-time | MacOS + iOS | Medium * | Medium * |
| Wispr Circulate | $15/mo | MacOS + iOS | ||
| Home windows | ||||
| Android | Very quick | Excessive | ||
| Willow Voice | $15/mo | MacOS + iOS | Very quick | Excessive |
✏️ One factor to notice: for instruments like Superwhisper and VoiceInk that run domestically, velocity depends upon which AI mannequin you select. Bigger fashions are usually extra correct however slower to course of.
How I am utilizing dictation in observe
Dictation rapidly turned a day by day a part of my work. Listed below are the methods it has been displaying up in my work in observe:
AI prompting
For probably the most half, I’ve stopped typing AI prompts when utilizing AI brokers for coding or in chat interfaces. This has been the most important unlock and time saver for me.
It is also made my prompts extra nuanced and detailed in comparison with once I sort them, which has improved the AI outputs consequently. When typing, I are likely to maintain issues transient out of comfort. When dictating, I naturally present extra context and rationalization.
Mind dumps and note-taking
When I’ve one thing on my thoughts, I sometimes dictate it rapidly to avoid wasting the context and are available again to it later. This may very well be venture concepts, reflections after conferences, or issues I am considering via.
I am not making an attempt to create polished writing in these moments – simply capturing ideas in a format I can revisit. The velocity of dictation makes this sensible in a approach that typing usually is not.
Messages and fast replies
I typically dictate shorter messages*,* too, although this varies. Dictation works significantly nicely for longer explanations the place I want to offer context. For very quick replies or messages with a number of formatting and hyperlinks, typing is normally nonetheless sooner.
Customized workflows
I’ve additionally developed some particular workflows round dictation. I’ve a couple of customized AI instructions in Raycast that course of dictation after which format it in several methods.
One use case I’ve discovered significantly helpful is for interviewing. We have been on a hiring push at Buffer, so I have been doing a variety of interviewing. After an interview, I’ve an AI command in Granola the place I dictate my ideas and impressions, and it constructions them as written suggestions for the interior notes. This has streamlined that course of considerably.
Reflections and proposals
I’ve realized loads all through this course of. Listed below are a couple of reflections and proposals if you happen to’re simply getting began with dictation or if you happen to’re on the lookout for a brand new device:
- Pace has a a lot larger influence on usability than I anticipated. These few seconds of ready for processing add up rapidly and break your movement in ways in which aren’t instantly apparent till you expertise a sooner device.
- I usually course of issues by speaking about them. Dictation instruments assist me assume via issues and discover the fitting phrases loads sooner than typing does. If you happen to’re somebody who thinks out loud, this is perhaps transformative on your workflow as nicely.
- This is not about changing typing totally. Completely different contexts name for various instruments. I’ve discovered dictation most respected for AI interactions, capturing ideas rapidly, and longer-form explanations. For fast replies, formatted textual content, or collaborative modifying, typing usually stays extra environment friendly.
- If you happen to’re inquisitive about making an attempt dictation instruments, I would suggest beginning with VoiceInk, given the low dedication. If you end up utilizing it usually and the constraints grow to be irritating, Wispr Circulate has been well worth the funding for day by day skilled use for me.
If you happen to give any of those instruments or ideas a strive, I would love to listen to from you. Be at liberty to achieve out on LinkedIn.
P.S. This put up was, in large half, dictated 😉
