Wispr Flow alternatives: local speech-to-text apps for macOS

Which macOS dictation tools suit someone who wants low cost, privacy and system-wide dictation? AR Labs carried out this survey in April 2026, using the subscription app Wispr Flow as the reference point, as part of our work on voice input to AI (see Voice UI). It compares consumer apps, open-source engines, cloud APIs and the built-in macOS options against vendor documentation, GitHub repositories and published reviews, with a confidence rating on each key claim.

Method: deep, multi-pass research, with confidence ratings per claim.

Summary

Wispr Flow charges $144/year for functionality that multiple free and one-time-purchase alternatives match or exceed. The macOS STT market in 2026 is dominated by Whisper-based local apps that process audio entirely on-device using Apple Silicon's Neural Engine. The three strongest alternatives for a user prioritising low cost, privacy, and system-wide dictation are:

1. open-wispr (free, MIT licence): installs in 5 minutes via a single curl command, works identically to Wispr Flow's core dictation loop
2. VoiceInk ($39.99 one-time, GPL source available): most actively maintained open-source app, 4,600+ GitHub stars, shipped v1.73 April 2026
3. Spokenly (free tier with local models + BYOK): developer-focused, Claude Code MCP integration, no subscription required for local use

For users wanting a polished experience with the most features, MacWhisper (€59 one-time) handles both dictation and batch transcription. Superwhisper ($849 lifetime) is the premium option for power users who need deep customisation and the most extensive mode system.

The landscape in six categories

Category 1: direct consumer apps (dedicated dictation)

These are the apps that compete directly with Wispr Flow. They sit in the menu bar, use a global hotkey, transcribe locally, and paste text wherever your cursor is.

Comparison matrix:

Key finding (high confidence): System-wide dictation (typing into any text field by holding a hotkey) is table stakes. Every app above supports it. The differentiators are pricing model, accuracy engine, and UX polish.

Key finding (high confidence): Wispr Flow's pricing is an outlier on the expensive end. Three independent sources (getvoibe.com, pikaseo.com, willowvoice.com) confirm $144/year positions it 40-80% more expensive than most alternatives.

Category 2: open-source STT engines

The underlying model layer powering most consumer apps. Relevant for users comfortable with CLI or building custom workflows.

Key finding (high confidence): whisper.cpp with CoreML acceleration is the dominant engine for local macOS apps. CoreML provides 3x speedup over CPU; combined Metal + CoreML achieves 8-12x. Every released consumer app uses whisper.cpp or WhisperKit under the hood.

Key finding (moderate confidence): Parakeet TDT 0.6B v2 achieves lower WER than Whisper on English (1.69% vs 2.2%) while using only 66MB working memory vs Whisper's 2GB+. Speed: 60-minute audio in 23 seconds on M1. VoiceInk and Spokenly both integrate it. For non-English use cases, Whisper remains the only serious option (99 languages vs Parakeet's 25 European languages).

Category 3: cloud STT APIs

Relevant if building custom workflows or using BYOK apps. Not a direct Wispr Flow replacement unless you write integration code.

Key finding (moderate confidence): For personal dictation at typical usage (~1-2 hours/month), OpenAI Whisper API costs ~$0.36-0.72/month, the cheapest cloud option. Deepgram Nova-3 is cheaper at scale and offers real-time streaming. Both are dramatically cheaper than Wispr Flow, but require technical integration work.

Evidence gap: No independent benchmark comparing cloud API accuracy to local whisper.cpp on the same audio corpus. Deepgram's WER figures come from Deepgram's own benchmarks, which carry bias risk.

Category 4: macOS built-in options

macOS Dictation (System Preferences → Keyboard → Dictation):

• On Apple Silicon Macs with macOS 13+: processed on-device, fast, private
• Global shortcut: double-tap Fn key (configurable)
• Works system-wide in any text field
• Accuracy: consistently reported inferior to Whisper-based apps for technical vocabulary, proper nouns, and accented speech
• Free, zero setup
• Limitation: no AI post-processing, no custom vocabulary for proper nouns beyond basic additions, no streaming preview

Apple Intelligence dictation (macOS Sequoia 15+):

• Available on M1 and newer with Apple Intelligence enabled
• Improvements in Sequoia focused more on Voice Control accessibility features than general dictation accuracy
• Automatic capitalisation of proper nouns added to custom vocabulary (Sequoia enhancement)
• User forums report ongoing quirks with formatting commands in 15.3.1
• Verdict: Free, private, good enough for casual use; not competitive with Whisper-based apps for power users

Voice Control (Accessibility → Voice Control):

• Distinct from Dictation: full computer control via voice, not just text input
• More comparable to Talon Voice than Wispr Flow
• Supports complex command structures

Key finding (high confidence): macOS built-in dictation is a viable free baseline but 3+ independent reviews (writingmate.ai, willowvoice.com, timingapp.com) consistently rate Whisper-based apps 15-40% more accurate, especially for technical content.

Category 5: self-hosted and local-first solutions

The spectrum from "turnkey app" to "DIY from raw whisper.cpp":

Turnkey (no coding):

• VoiceInk, open-wispr, BetterDictation, Spokenly free tier (see Category 1)

Semi-technical (1-2 commands):

• open-wispr: single curl command installs everything including whisper.cpp via Homebrew
• local-whisper: bash installer, requires Hammerspoon for hotkey management
• whisper-mac (Explosion-Scratch): electron app, supports whisper.cpp + Vosk + macOS Speech

DIY / developer tools:

• whisper.cpp + custom hotkey script
• WhisperKit CLI: brew install whisperkit-cli then argmax-cli transcribe --stream
• local-stt-mcp (SmartLittleApps): MCP server for whisper.cpp, optimised for Apple Silicon, 15x+ realtime speed

Key finding (high confidence): open-wispr is the most accessible path to zero-cost system-wide dictation. Single curl command, no Homebrew manual setup, auto-downloads model, installs macOS service. Only 103 GitHub stars but shipped v0.34.0 in April 2026, so it is actively maintained.

Category 6: hybrid solutions

Apps offering both local and cloud processing, user-selectable:

• Spokenly: Local Whisper/Parakeet models free; BYOK (OpenAI, Deepgram, Groq) free; Pro $9.99/mo for managed cloud
• Superwhisper: Free tier uses small local models; Pro ($85/yr or $849 lifetime) unlocks both cloud and larger local models
• OpenWhispr: Free tier has 2,000 words/week cloud + unlimited local; Pro $6.67/mo for unlimited cloud
• whisper-mac: Supports whisper.cpp (local), Vosk (local), macOS Speech (local), or cloud; user picks per session

Recommendations by profile

"I want Wispr Flow behaviour for free, right now"

open-wispr. Single command install:

curl -fsSL https://raw.githubusercontent.com/human37/open-wispr/main/scripts/install.sh | bash

Hold Globe key, speak, release, and text appears at the cursor in any app. Runs whisper.cpp with Metal acceleration. MIT licence, no telemetry, no accounts. Works on macOS 13+ with Apple Silicon.

Limitation: English-only models by default (use large-v3-turbo for multilingual). 103 GitHub stars, so a small community, but v0.34.0 shipped April 2026.

"I want the best free/open-source app with polish"

VoiceInk. 4,600 GitHub stars, v1.73 shipped April 10 2026, GPL v3. $39.99 for the binary with auto-updates; build from source for free (requires Xcode). System-wide hotkey, Power Mode (per-app settings), 100+ languages, Parakeet model available for English. macOS 14.4+ required.

If cost is the constraint: clone the repo and build in Xcode. The binary purchase funds ongoing development.

"I want free and want BYOK flexibility"

Spokenly. Free tier includes unlimited local Whisper/Parakeet models and full BYOK support (OpenAI, Deepgram, Groq, and others). If you already have an OpenAI API key from other tools, Spokenly gives you cloud-quality dictation at zero marginal cost. Pro is $9.99/month for managed cloud + iOS. Also ships an MCP server for Claude Code voice input.

"I want one polished app, pay once, done"

VoiceInk at $39.99 is the best value. MacWhisper at €59 if you also need batch transcription and audio file processing. BetterDictation at $39 is a close competitor requiring Apple Silicon only.

"I want maximum features and can pay once"

Superwhisper lifetime at $849. Premium pricing for the most mature mode system, context awareness (reads selected text, clipboard, app context), and Macrowhisper automation. Note: community feedback in 2025-2026 flags the developer reducing settings granularity and changing interface; "identity crisis" described by one experienced reviewer. Consider VoiceInk or Spokenly if Superwhisper's direction concerns you.

"I code and want voice commands, not just dictation"

Talon Voice (free public beta). Full programmable command system: navigate files, execute git commands, control cursor. Not a dictation app: designed for short voice commands that trigger actions, not continuous prose transcription. Requires 2-4 weeks to build fluency. Suited to technical users comfortable with Python config scripts.

Confidence ratings by claim

Evidence gaps

macOS 16 / future Apple Intelligence dictation: Apple has not announced dictation-specific improvements for the next macOS cycle. Current indications from Sequoia suggest accessibility features are the focus, not general dictation accuracy.

Independent accuracy benchmarks: Most WER figures come from the apps themselves or affiliated reviews. No recent independent corpus test comparing all six major consumer apps on the same audio. The macparakeet.com Parakeet figures are from the developers of that implementation.

Wispr Flow local processing scope: Wispr Flow's marketing describes "on-device processing" but doesn't clearly document which features require cloud. Multiple comparison sites classify it as cloud-primary, which partially explains its higher pricing.

Sources

• Wispr Flow pricing: https://wisprflow.ai/pricing
• VoiceInk GitHub: https://github.com/Beingpax/VoiceInk
• open-wispr GitHub: https://github.com/human37/open-wispr
• Spokenly pricing: https://spokenly.app/pricing
• Superwhisper website: https://superwhisper.com/
• MacWhisper review (daveswift.com): https://daveswift.com/macwhisper/
• Parakeet vs Whisper benchmarks: https://macparakeet.com/blog/whisper-to-parakeet-neural-engine/
• Apple Silicon Whisper benchmarks: https://www.voicci.com/blog/apple-silicon-whisper-performance.html
• AI dictation app differentiators analysis: https://afadingthought.substack.com/p/best-ai-dictation-tools-for-mac
• Superwhisper vs VoiceInk: https://blog.apps.deals/2025-04-23-superwhisper-vs-voiceink
• WhisperKit GitHub: https://github.com/argmaxinc/WhisperKit
• BetterDictation review: https://betterdictation.com/
• Deepgram pricing benchmarks: https://deepgram.com/learn/speech-to-text-api-pricing-breakdown-2025
• Cloud API comparison: https://vocafuse.com/blog/best-speech-to-text-api-comparison-2025/