Why? Because I write slower than I think.

To remedy this, I gave a shot at voice dictation tools last August.

Putting down your ideas has never been more fun.

But whoever I ask to try them faces the same set of issues.

“Which one should I use?”

“Is the paid subscription even worth it?”

“Only 2000 words per week doesn’t cut it.”

So they go back to typing.

Traditional media doesn’t help either — NYTimes and PCMag have some horrible suggestions.

So I tested a bunch of tools to find the answers for you:

(Trust me, with every developer building their version of these apps, you’ll be okay limiting your options to these.)

TL;DR

1. MacWhisper: Best overall (especially on Mac)

2. WisprFlow: Best premium experience

3. Superwhisper: Most powerful / complete

4. Google AI Edge Eloquent: Best for mobile

Everything else: Meh / niche

0/ Context Before We Start

A few things most people don’t realise about voice dictation apps:

• Almost all tools today are very good at detecting English.

• Good models even do live translations and transliterations.

• Almost all apps can run on local models (no internet needed)

• Local = more privacy, but uses your RAM

• Most tools work like this:
  → Press the Fn key → speak → text appears

• Some apps offer features like:

  • auto-formatting, grammar correction, etc.

  • custom dictionary on the most used words, like names and your industry lingo that LLMs might get wrong

  • tones — legal, formal, casual, etc.

Note: We’ve previously partnered with Superwhisper — but this review isn’t driven or motivated by that.

1/ MacWhisper: The Default Choice

Website: MacWhisper

Works on MacBooks only.

Score: 10/10

If you want something that just works, this is it.

The Good

• Free for core feature of voice dictation

• One-time payments, no subscriptions

• Extremely reliable for everyday dictation

The Bad

• Mac-only

• No. of fancy features

• UI is functional, not pretty

One-Sentence Verdict

If you want something that just works, this is it.

2/ SuperWhisper: Most Complete!

Website: Superwhisper

Works on: Desktops (Mac, Windows) and iPhone

Score: 10/10

This feels like it was built by someone who actually uses dictation daily.

The Good

• Baked-in tones modes — formal, casual, etc.

• Internal vocabulary: names, acronyms, slang

• Multi-lingual support

• Choose your own LLM models

• Integrates with other AI apps

• More economical than WisprFlow:

  • The basic features are free forever

  • Has a buy once, own forever option

The Bad

• No android app as of today.

One-Sentence Verdict

If you want control + features + flexibility → this is the best tool today.

3/ Wispr Flow: Most Premium

Website: Wisprflow AI

Works on all desktops (MacOS, Windows), and all phones (Android/iOS)

Score: 9/10

This feels like the Apple product of dictation tools.

Clean. Polished. Magical.

The Good

The features in the app make the interaction magical.

Feature-packed with auto-formatting, an internal dictionary, snippet insertion, automatic-cleanup and multiple tones and styles, it stands

You’ll get these features elsewhere, too, but it won’t feel the same.

The Bad

• 2000 words per week on desktop and 1000 words per week on iOS is a real limitation

• Premium pricing at $12/month and no lifetime pricing

One-Sentence Verdict

If money isn’t a constraint and you want the smoothest UX → this is it.

4/ Google AI Edge Eloquent: Best for iOS

Website: Google AI Edge Eloquent

Available as: iOS app only

Score: 8/10

Simple, local-first dictation for iPhone users.

The Good

This is available for free for all iOS users.

The models are stored on-device locally, or you can use the cloud models too.

The Bad

• iOS only

• Basic UI

One-Sentence Verdict

If you just want something free on your iPhone, use this.

5/ Monologue: Only for Devs

Website: Monologue

Works on: Apple ecosystem only (iPhone, Apple Watch, Macs, iPad)

Score: 5/10

Beautiful product. Niche audience.

The Good

• One of the smoothest and most fun onboarding experiences of all the products.

• Fun interface, micro-gamifications

The Bad

• Apple ecosystem only

• Offline local models are paid too.

• Only 1000 lifetime words in the free plan. 🤦‍♂️

One-Sentence Verdict

Interesting product if you’re a dev with cash to throw.

6/ SpeakMac: Simple & Honest

Website: SpeakMac

Available as a macOS app only

Score: 5/10

Basic voice dictation tool for MacOS users.

The Good

$19 for a lifetime licence.

Models are saved locally.

The Bad

• No free plan

• No iOS, Windows or Android app

• Unlike SuperWhisper and MacWhisper no flexibility in model choice

• The local model is ~600 MB big — this will hang your machine if you use more than 1 browser window.

One-Sentence Verdict

Cheap, but you’ll outgrow it quickly.

7/ Willow Voice: 🤦‍♂️ Hard to Recommend

Website: Willow Voice

Works on: macOS, iPhone and Windows

Score: 4/10

As basic as these tools come, tone configuration and internal vocabulary.

The Bad

• 2000 free words per week on the free plan

• Offline models available via Pro subscription only 🤦‍♂️

• At $12/month, this is at par with WisprFlow but with fewer features. 🤦

One-Sentence Verdict

There are strictly better options for the price, aka WisprFlow!

8/ Aqua Voice: 🤦‍♂️🤦‍♂️🤦‍♂️ Skip

Website: Aqua Voice

Works on: All desktops (Macs/Windows) and iPhones

Score: 3/10

Has all the usual features: custom instructions on tone, custom dictionary, snippet insertion and auto formatting.

The Bad

• No feature for local models

• Only custom models by the Aqua team

• Limited to JUST 1000 words in the free plan

• Can only log in with a Google or Apple account.

In Closing

Typing is a bottleneck.

And for the first time, voice dictation is actually good enough to replace it — at least partially.

If you’re still feeling stuck, here’s the simple way to decide:

• Beginner? MacWhisper

• Want the best deal? Superwhisper

• Want premium UX/features? WisprFlow

• Mobile-first? Google AI Edge Eloquent

Hope this post helps you navigate the tricky waters of choosing an AI tool that best fulfils your use cases.

Until next time.

Auf Wiedersehen!

Medvi reminded me of how in every tech cycle we get new pioneers in customer acquisition. Be it Amazon never turning a profit or WhatsApp not charging users a dime for new users (in India).

Not illegal, but definitely not moral.

All these stories become templates for the next set of founders on cracking markets where cracking seemed impossible.

I think, to that end, Medvi just showed the future of AI-native distribution.

In fact, anyone doing $401M revenue in their first year ought to be studied.

Especially when they’re doing it solo!

So, while the world debates what is right and what is not, we’ll discuss Medvi’s playbook and how they cracked the already crowded GLP-1 market.

0/ About Medvi

Medvi is a telehealth startup created by Matthew Gallagher that sells GLP-1 drugs. These drugs help reduce body weight by tricking the body into feeling satiated and full for a longer time.

But it’s not a drugmaker or a clinic-chain owner.

Medvi’s business model is rather simple.

It partners with platforms like CareValidate and OpenLoop that handle doctors, pharmacies, prescriptions, shipping and everything in between.

Medvi just gets them customers via their distribution.

And the result?

$1.8B ARR in 2026.

$410M revenue, and

$63M in profits in 2025.

All this on a budget of $20K.

Now, these numbers make more sense when you realise that the solopreneur in question is a veteran entrepreneur. He ran Watchgang, a watch subscription company, before this with varying success.

Running such an e-com venture, that too for 7 years, teaches you a thing or two about what works and what doesn’t.

All the lessons accrued at WatchGang became a template for Medvi, and today they’re our masterclass in AI-native growth.

1/ Zero-Day Distribution Hack

Before the AI personas. Before the automated support stack. Before any of the AI workflows, the one foundational strategy that resulted in Medvi’s blitz success:

outsourcing distribution; instead of building one.

In any venture, a marketing team would hire creators. Negotiate rates. Build a content calendar. At $2,000 per reel for mediocre reach, they’d be looking at a seven-figure budget before even a single sale.

Then there’s SEO, content, personal branding, newsletters ;-; … all led by teams of people.

This is templatised now.

This is pretty evident by their peers in the same segment – Hims and Hers – which do $2.5B annually with ~2500 employees.

But Medvi chose differently.

They chose the oldest unlock in consumer internet: paying users to bring others to their network.

Think of PayPal paying $10 to sign up and $10 to refer or Dropbox offering extra storage for each new customer referred.

They built an affiliate programme generous enough to turn its distributors into missionaries.

And when Medvi’s affiliates were making real money, they flocked.

This reportedly resulted in some 800 fake-doctor profiles on Facebook pushing GPT-1 drugs towards this one company:

The lesson here is clear.

If your unit economics supports it, make your affiliates richer than your competitors will, and the distribution problem solves itself.

2/ The AI First Way

For an operation like Medvi’s, you don’t just stop at distribution.

Anyone who’s scaled beyond 10 employees knows this: once distribution is working, infrastructure and operations become the bottleneck.

This includes, your website, customer support, creatives, workflows, etc.

All of which need to run smoothly, 24/7.

So how does a solopreneur handle this?

By being AI-first everywhere.

• Claude / Replit / Cursor for the website

• ElevenLabs for inbound voice

• CareValidate and OpenLoop for clinical and regulatory layers

• Contract engineers for specific builds — brought in and rotated out

• Agentic workflows to automate all hygiene tasks

No org chart. Just a routing map.

This is what AI leverage actually looks like when someone means it.

Every function that traditionally required a hire was converted into either:

• a vendor relationship, or

• an automated workflow

Support, compliance, content, customer communication — all orchestrated, none owned.

This case highlights a deeper shift:

When doing becomes cheap, value moves to orchestration.

Most founders are still building in-house what they could orchestrate externally in a week — because building feels like progress, while routing feels like admitting you don’t have a team.

Medvi didn’t have that ego.

That was their advantage.

Shameless Plug: We’ve been running AI audits for some of our readers — especially businesses here in India and the US — and have seen strong results.

So, we’re opening this up to everyone.

It’s free (if you refer another business), and you’ll get a clear, actionable strategy on where AI can actually move the needle in your company.

If this sounds interesting, you can share your details here.

3/ Constraints are a blessing!

In 2011, The Lean Startup became a rage across startup circles worldwide.

Instead of building large teams — which was the norm at the time — it argued that companies should aim to do more with less to improve their chances of success.

I read it in 2014, and the lessons were already loosely getting adopted.

Today, that thinking is standard.

But, in Medvi’s case, it was dialled up to 12.

Matthew had previously run a 60-person company and knew the hidden costs of maintaining a team.

Headcount becomes the default solution to every problem. Tools never get a real chance to prove themselves, because there’s always a person available to do things the old way.

This is more evident by his op-ed on NBC News .

That’s why Medvi’s $20K starting point wasn’t a limitation but an advantage.

More capital might have prevented AI-native thinking altogether.

The low budget removed that option.

Matthew had to automate — or stall.

And automated he did. To $1.8B before finally bringing in his first employee – his younger brother.

In Closing

For many, Medvi isn’t the “pure” 1-person AI company to do $1B ARR.

Some founders will look at this story and see only a white-collar online scam.

But the real takeaway isn’t the controversy.

It’s the playbook:

• distribution as the core moat

• affiliates as a growth engine

• AI as an orchestration layer, not just a tool

These ideas are already seeping into how modern companies will be built.

You don’t have to agree with how it was done.

But it’s worth paying attention to what was done.

And we must appreciate him for showing us the way!

PS: This is the first edition of AI Case Studies by This Week in AI. We’re calling this Transfirm.

Going forward, you’ll receive two formats — alternating each week:

• our regular This Week in AI newsletter

• and a deep-dive case study on how companies are actually using AI to win

Hope you like this initiative!