Voice Dictation & Transcription

Hands-on-the-keyboard voice typing — dictate Slack messages, write emails, draft notes. Distinct from podcast transcription (see transcription). For broader AI assistants see prod-ai-productivity-assistants; for accessibility / motor-disability voice control see notes below.

Mac

• ★ Wispr Flow — paid (subscription); the breakout 2024–26 voice-typing app. Hold a hotkey, dictate naturally with disfluencies, get clean punctuated text inserted in any app. Closed-source but transformative.
• ★ Apple Dictation — free, built-in; double-Fn (or Edit > Start Dictation); offline since 2022; surprisingly good in macOS Sequoia with Apple Intelligence.
• MacWhisper — paid one-time (~$60); local Whisper.cpp; transcribe audio files OR live-dictate; offline; loved by privacy-first users.
• Whisper Transcription (Aiko-shape apps) — paid Mac App Store; offline Whisper.
• Whisper Memos — paid; iOS/Mac voice memo + transcript; pretty.
• Aiko — paid one-time Mac; offline transcription.
• Superwhisper — paid; the most direct Wispr Flow competitor; local + cloud models; bring-your-own-API-key option.
• VoiceInk — free OSS; new MacWhisper-style local-first dictation — emerging 2024.

Windows

• Windows Speech Recognition — free, built-in (Win+H); deeply improved with Voice Access in Win11.
• WhisperWriter — free OSS; local Whisper voice typing.
• Talon Voice — free + paid Pro; programmable voice control; dictation + commands ("scratch that", "select word"); accessibility-leaning power tool.
• Dragon Professional (Nuance) — paid; the medical / legal flagship; expensive; very accurate for domain language.
• Superwhisper — paid; cross-platform now.

Linux

• nerd-dictation — free OSS; Vosk / Whisper based.
• Whisper.cpp — free OSS; raw model + your own scripting.
• Numen — free OSS; voice command + dictation.
• Talon Voice — free + paid Pro; works on Linux.
• dictation tools are weakest on Linux — most users self-build with whisper.cpp.

iOS / Android

• ★ iOS Dictation — free, built-in; on-device since iOS 16; tap mic on keyboard.
• iOS Voice Control — free, built-in (Settings > Accessibility); full voice navigation; powerful for handsfree.
• Apple Voice Memos — free; transcription on iOS 18+.
• Otter.ai — paid + free; meeting transcription; live notes.
• Just Press Record — paid + free; one-tap voice memo with auto-transcript; Apple Watch.
• Google Assistant Voice Typing — free, built-in on Pixel/Android.
• Whisper iPhone apps — Whisper Memos, etc.

Underlying models / SDKs

• Whisper (OpenAI) — open weights; the de-facto offline transcription model; many forks (whisper.cpp, faster-whisper, distil-whisper).
• Deepgram / AssemblyAI — hosted ASR APIs — see transcription.
• Vosk — free OSS; lightweight offline ASR.
• NVIDIA Parakeet / Canary — newer open ASR models with strong English accuracy.

Programmable voice control (accessibility)

• ★ Talon Voice — free + paid Pro; macOS/Win/Linux; voice + eye-tracking + foot-pedal; the gold standard for hands-free coding for users with RSI or motor disabilities. Strong Python / Cursorless community for code editing by voice.
• Cursorless — free OSS; works inside VS Code with Talon; navigate code by voice with hat-marked targets.
• Voice Control (macOS / iOS Accessibility) — free, built-in; navigate the OS by voice; less programmable than Talon.
• Dragon NaturallySpeaking (Win) — paid; medical / legal accuracy.

What "voice typing in 2026" actually means

• Disfluencies handled — "uh, no wait, scratch that" is removed by Wispr Flow and similar; you can talk like a human.
• Punctuation auto-inserted — Whisper-class models predict punctuation from context.
• App-aware insertion — paste into any app, not a special "voice input" field.
• Live transcription vs batch transcription — Wispr / Superwhisper are live; MacWhisper / Aiko also do batch.
• Custom vocabulary — names, jargon, code identifiers; Wispr / Talon / Dragon support this.
• AI-cleanup — pass the transcript through GPT-4 / Claude before insertion; Wispr does this; Superwhisper has it.

Privacy reality

• ★ Apple Dictation (offline) — runs on-device on Apple silicon; nothing leaves your Mac.
• MacWhisper / Superwhisper local mode / VoiceInk / Whisper.cpp / Aiko — local; private.
• Wispr Flow — cloud-based; transcripts processed on their servers — read the privacy policy if dictating sensitive info.
• Otter.ai / Fireflies / Granola — cloud + AI summaries; not for confidential meetings unless on a paid plan with retention controls.

Pricing reality check

• Apple / Win / Google built-in dictation are free and good.
• Wispr Flow ~$15/mo — the polished hosted experience.
• MacWhisper / Aiko / VoiceInk are one-time or free — local + private.
• Talon Voice is free with paid Pro tier (community supported); Cursorless is free OSS.
• Dragon Professional is ~$700 — niche to medical/legal.
• Superwhisper ~$10/mo or BYOK — closest to Wispr Flow with local options.

Patterns that actually work

• ★ Bind dictation to a hold-to-talk hotkey. Walkie-talkie shape > toggle on / off; lower error rate.
• Train your model on your own voice if the tool supports it (Dragon, Talon).
• Local Whisper on Apple silicon is shockingly good — no excuse to send dictation to the cloud unless you want the AI cleanup that Wispr provides.
• Pair with prod-text-expansion — short, common phrases via expansion; long-form via voice.

Pick this if…

• Default Mac voice typing in 2026, willing to pay: Wispr Flow.
• Default Mac, free, on-device: Apple Dictation.
• Mac private + offline + audio file transcription: MacWhisper.
• Mac local Wispr-shape: Superwhisper or VoiceInk.
• Default Win, free: Win+H dictation.
• Win OSS local: WhisperWriter.
• Programmable voice control (RSI / accessibility): Talon Voice.
• Highest accuracy domain (medical/legal): Dragon.
• Phone: built-in (iOS / Pixel) — third-party rarely beats them.