XP Daily Plain comparisons of personal productivity tools

How to dictate code comments and technical specs for software engineers in 2026

A plain guide for software engineers to dictate code comments and technical specs on Windows, with steps, tools and the mistakes to avoid.

By XP Daily ยท

XP Daily is published by the makers of DaysAid, QuickDictate and YTSort, which are compared here alongside other tools.

Install a speech-to-text tool on a 64-bit Windows PC, pick a recognition provider or an offline model, and dictate comments and specs into the editor, terminal or chat window that has focus in short passes. Read every dictated line back before you keep it.

OptionIts own feeCard fee
Writing by handnot publishednot published
QuickDictatefree for personal and other noncommercial use [1]not published

What you need before you start

Dictating comments and technical specs takes a few things ready first.

Set the keys and vocabulary before you dictate a real file, so the first pass is not spent fixing settings.

Step by step

These steps work whatever dictation setup you choose.

  1. Decide what to dictate. Pick comments, a design note or a set of technical specs, and keep each pass to one kind.
  2. Open the window you want the text in. What you say goes into whichever window has focus. [1]
  3. Set your keys. Hold a key while you talk, or press it once to start and once to stop, and you can change both keys. [1]
  4. Add vocabulary for your project. List words and phrases that are sent to the provider to help it recognise them, and add replacements that fix spellings such as "Chat GPT" to "ChatGPT". [1]
  5. Choose where recognition runs. Pick a cloud provider and that provider bills you directly, or choose Local with a model such as Whisper Large v3 Turbo so no internet connection or API key is needed. [1][1]
  6. Dictate a line and read it back. Say the comment, stop, and check the window. Watch for terms the recogniser guessed.
  7. Fix what you mis-speak. Say "scratch that" to discard the text you just dictated instead of inserting it. [1]
  8. Use the history if you need it. You can browse and filter your last 50 dictations in Settings, and the history survives a restart or self-update. [1]
  9. Review the whole file. Read every dictated line once before you commit it, because speech recognition can change the terms you rely on.
  10. Keep the review habit. A quick look after each pass beats cleaning up a whole file later.

Tools that can do it

You do not need a special tool to write comments and specs. Here are options to consider.

Writing by hand. Type comments and specs into the editor yourself. Nothing is published here about price because it depends on the tools you already own.

QuickDictate. QuickDictate is made by the team that publishes this site. It is a Windows program that types what you say into whichever window you are using. [1] You hold a key while you talk, or press it once to start and once to stop, and you can change both keys. [1] What you say is typed into whichever window has focus, such as an editor, chat box, terminal or web form. [1]

You choose the recognition source. Use ElevenLabs, Deepgram, OpenAI, AssemblyAI, DashScope or Google, and that provider bills you directly. [1] With five of the cloud providers, words appear while you are still talking. [1] Choose Local in Settings and install a speech model such as Whisper Large v3 Turbo, and dictation needs no internet connection or API key. [1]

Profiles can change the provider, language, vocabulary and formatting depending on which app is in focus. [1] You can list words and phrases that are sent to the provider to help it recognise them, and a replacement table fixes spellings such as "Chat GPT" to "ChatGPT". [1] While you dictate, other apps' audio is muted or turned down, and it goes back to its old level when you stop. [1] You can browse and filter your last 50 dictations in Settings, and the history survives a restart or self-update. [1]

QuickDictate is free for personal and other noncommercial use, with no subscription or account, and cloud providers bill you separately for their service. [1] From v0.9.0 on it is source-available under PolyForm Noncommercial 1.0.0: free for personal and other noncommercial use, and commercial use needs a separate licence; releases up to v0.8.0 were MIT. [1] It runs only on 64-bit Windows 10 and 11. [1] It is delivered as an .exe file on GitHub, with a built-in updater that works silently, and it can also be built from source. [1] If you sign in, preferences such as hotkeys and language can be synced between PCs, while API keys, audio and transcripts are not part of that sync. [1]

Common mistakes to avoid

Dictating long paragraphs in one pass. Speech recognition drifts on technical text, so one line at a time is easier to check.

Skipping the read-back. Every dictated line can carry a wrong term, so re-read it before you keep it.

Letting project terms get mangled. Add the words you use often so the provider hears them, and add replacements for spellings such as "Chat GPT". [1]

Working in a noisy spot. Other apps' audio can crowd your speech; while you dictate, other apps' audio is muted or turned down. [1]

Forgetting the history. You can browse and filter your last 50 dictations in Settings, and the history survives a restart or self-update. [1]

Committing before checking. Merge dictated text into the file only after you have read each line and fixed the terms.

Frequently asked questions

How to dictate code comments with QuickDictate?

Open the editor window, hold or tap the key you chose while you speak, and the words appear in the focused window. [1] Choose a provider such as ElevenLabs or OpenAI, or Local with a model like Whisper Large v3 Turbo, then say each comment and read it back. [1][1]

Can QuickDictate work offline for company code?

Yes. Choose Local in Settings and install a speech model such as Whisper Large v3 Turbo, and dictation needs no internet connection or API key, which suits code that cannot leave the building. [1] Cloud providers send your speech to their services, so offline is the choice when a project stays on the machine.

Does QuickDictate cost anything for a personal project?

It is free for personal and other noncommercial use, with no subscription or account. [1] Cloud providers bill you directly for their service. [1] Commercial use needs a separate licence from v0.9.0 on. [1] Releases up to v0.8.0 were MIT. [1]

Can QuickDictate fix how terms like Chat GPT come out?

Yes. You can list words and phrases that are sent to the provider to help it recognise them, and add a replacement table that fixes spellings such as "Chat GPT" to "ChatGPT". [1] Per-app profiles can change the provider, language, vocabulary and formatting depending on which app is in focus. [1]

Should software engineers always dictate their specs?

Not always. Speech recognition can change technical terms, so short comments and spec lines work better than long paragraphs, and every line still needs a read-back before it is committed. When exact spelling matters most, typing remains the reliable option.

QuickDictate is free for personal and other noncommercial use, with no subscription or account. Cloud providers bill you separately for their service. [1]

Try QuickDictate

Sources

  1. QuickDictate's own description of itself (the product of this site's publisher)

Related comparisons