Blog

Notes on talking to your computer.

Guides on voice dictation, AI prompting, and keeping your speech private, from the team building SpeakoFlow.

Seven cleanup models ranked by overall score with the do-nothing baseline fourth as an empty dashed bar and four models below it.
Release

SpeakoFlow Mini: a local dictation cleanup model that knows when to change nothing

An 833 MB cleanup model that runs offline on a CPU, returns already-correct transcripts untouched 92.6% of the time, and scores 70.7% where a program that changes nothing scores 50.0%.

10 min read
Read the release notes
One model before and after fine-tuning, with overall score, restraint and edit accuracy side by side against the do-nothing line.
Guide

How to clean up dictated text: rules for the patterns, a small model for the judgement

Deleting the word um is a table lookup, and deciding whether no, Friday cancels Thursday is not. Two stages, why the order is not negotiable, and how to measure whether your cleanup is helping at all.

12 min read
Read the guide
Wispr Flow marked as having no Linux build beside SpeakoFlow's AppImage, .deb and Arch package, with the GitHub star counts of six tools that do run on Linux.
Roundup

Best dictation software for Linux in 2026: the paid apps don't ship here

Six tools that genuinely run on Linux, checked against their own repositories today, and the five popular apps that have no Linux build at all.

11 min read
Read the roundup
Wispr Flow's three published plans beside the annual cost they add up to, next to zero.
Pricing

Wispr Flow pricing in 2026: the free tier, the real annual cost, and the $0 alternative

Free is $0 with a 2,000 word weekly cap on desktop, Pro is $15 per user per month or $12 billed annually, and Enterprise carries no published figure.

11 min read
Read the breakdown
Speech travelling to a vendor server and back on one path, and staying on the machine on the other.
Answer

Does Wispr Flow work offline? No, and its own docs say why

Wispr Flow transcribes every dictation on its servers, which its own Data Controls page states outright, so with the network off the desktop app tells you before you even start speaking.

10 min read
Read the answer
A cost comparison: Wispr Flow Pro at $180 a year billed monthly or $144 billed annually, against SpeakoFlow, Handy and Buzz at zero.
Comparison

Best free and open-source Wispr Flow alternatives (2026)

Wispr Flow's free tier caps you at 2,000 words a week. Here are the free and open-source tools that do the same job on your own machine, compared on price, platform, and privacy.

9 min read
Read the comparison
A spoken question about an on-screen error, and the assistant answering it with the port number and the command to fix it.
Guide

How to get a Jarvis on your own computer

Search for one and you get a dozen Python scripts you cannot run. Which parts of the idea actually work today, and how to set them up without writing code.

9 min read
Read the guide
SpeakoFlow and Wispr Flow side by side: transcription on your device at zero against transcription on their servers at $15 per user per month.
Comparison

SpeakoFlow vs Wispr Flow: free and local vs cloud

Both turn your voice into text in any app. The real difference is where your voice goes, what it costs, and whether you can see the code.

8 min read
Read the comparison
The same request typed as four words and spoken as five lines, with the spoken version carrying the file, the cause and the test to add.
Guide

Voice dictation for AI prompting: talk to ChatGPT, Cursor & Claude Code

You think in full paragraphs but type in fragments, and AI tools reward the paragraphs. Here is how to close that gap with your voice, without sending it to anyone.

7 min read
Read the guide