Short answer: AICHE runs on macOS, Windows, Linux, iPhone, iPad, Apple Watch, Android, Chrome, Obsidian, and via REST API. Willow Voice runs on macOS, Windows, iPhone, and Android. If you live on Linux, want a public API, or need a team plan without a 3-seat minimum, AICHE is the answer. If you want a polished Mac/iOS dictation app with a "style memory" feature and you don't mind paying $144/year, Willow is fine.
What Willow Voice is selling
Willow Voice is a YC X25-backed dictation startup from two Stanford dropouts (Allan Guo and Lawrence Liu), launched in March 2025. The pitch from their homepage: "Stop typing. Start writing 5x faster with voice." Their YC profile calls it "the voice interface replacing your keyboard."
The core feature they emphasize is style memory: Willow learns your writing voice and adapts tone per app. Casual in Slack, professional in Gmail, code-aware in Cursor. They publish a 200ms transcription-to-text latency figure on their own benchmark post (no independent third-party re-measurement located as of May 2026), advertise an "AI Mode" that expands brief verbal phrases into full messages, and ship voice commands like "dash," "new line," "bullet point." Their iOS app pairs a dictation app with a dedicated AI Keyboard for quick text edits.
The marketing surface (willowvoice.com) advertises "SOC 2, HIPAA, zero data retention, and privacy mode" as compliance posture. Per Willow's published pricing page as accessed May 2026, the SOC 2 + HIPAA + zero-retention combination - alongside MSA & DPA, SSO/SAML, and advanced admin controls - is listed under the Enterprise tier (custom pricing, billed annually). The Individual ($12/mo annual) and Team ($10/mo per seat, 3-seat minimum) tiers as shown on that page do not include the same contractual commitments. They ship fast: Windows landed January 2026, Cursor support February 2026, Teams plan March 2026.
The feature comparison
| Capability | AICHE | Willow Voice |
|---|---|---|
| Platforms | macOS, Windows, Linux, iPhone, iPad, Apple Watch, Android, Chrome extension, Obsidian plugin, REST API | macOS, Windows, iPhone (+ AI Keyboard), Android |
| Linux | Yes (.deb, .rpm, AppImage, Flatpak) | No |
| Apple Watch | Yes (record from wrist, syncs via iPhone) | No |
| Chrome extension | Yes | No |
| Obsidian plugin | Yes (Community Plugins) | No |
| Free tier | 7-day full trial, no credit card | 2,000 words/week (resets weekly), no credit card |
| Entry paid price | $3.99/mo (Personal annual, $47.99/yr) | $12/mo (Individual annual, $144/yr) |
| Pro / power tier | $8.33/mo (Pro annual, $99.99/yr) | None advertised between Individual and Team |
| Team plan | Pro tier includes team management, no listed seat minimum | $10/mo per seat, 3-seat minimum |
| Public REST API | Yes (Pro) | Not advertised |
| Software Development profile | Yes (Pro) - tuned for code identifiers, CLI flags, library names | Cursor/VS Code integration, no advertised code-tuned recognition profile |
| Custom vocabulary | Yes, 50 entries synced across platforms | Yes (personalization + shared team dictionary on Team tier) |
| Voice commands ("new line", "dash") | No (AICHE inserts plain transcribed text) | Yes |
| Style memory (tone per app) | No | Yes (their headline feature) |
| Hotkey model | Global hotkey, toggle on/off (⌃+⌥+R Mac, Ctrl+Alt+R Windows/Linux) |
Global hotkey |
| Recording length cap | None | "Longer dictation sessions" on Individual |
| Voice input languages | 99 | 100+ claimed (same underlying Whisper engine, comparable coverage) |
| Mobile UI languages | 28 | Not specified |
| Auto-translate to English | Yes, on every platform | Cited by one user as occasionally translating when not intended |
| Cold start to recording | Sub-100ms | 200ms transcription-to-text (different metric) |
| 15 minutes of audio to text | ~3 seconds | Not published |
| Audio retention | within 1 second after processing, named cloud provider (Groq) | "Zero data retention" gated to Enterprise tier per pricing page; default tiers retain "recognized dictated text (anonymized)" under Opt-In mode or operational telemetry under Private mode |
| E2EE sync | Yes (AES-256-GCM, Argon2id key derivation, opt-in) | Not advertised |
| Offline mode | No (cloud transcription, with crash-proof local Later queue) | Optional Offline Mode on Mac + iOS (local model fallback) |
| HIPAA / SOC 2 | Not advertised | Advertised on homepage; gated to Enterprise tier on pricing page |
| Crash-proof save | Yes (Later queue) | Not advertised |
| Privacy posture | Cloud transcription, named provider, immediate audio purge after processing, E2EE sync, no keystroke logging, no clipboard monitoring, no window-title transmission | Cloud-first default with optional local fallback on Mac/iOS. Their published privacy policy (last updated 2025-04-30) does not name specific subprocessors or specify a server-side audio retention duration |
Where Willow Voice wins
A few places, and they are real:
Style memory is a genuine feature. This matches what Willow's marketing describes - their core differentiator is adapting tone per app. AICHE has AI cleanup (filler removal, punctuation, paragraph breaks) but does not learn a per-app voice. If you write in a sharply different register in Slack vs. Gmail vs. a code review, and you want the dictation app to track that, Willow does something AICHE does not.
iOS AI Keyboard. Willow ships a dedicated iOS keyboard app for quick edits, sitting alongside their dictation app. AICHE's iPhone experience is capture-first (you record a memo, the app holds it). If you want iPhone-keyboard-style inline dictation that lives inside the iOS keyboard layer, Willow has that and AICHE doesn't.
Voice commands. "Dash," "new line," "bullet point" - Willow supports literal voice instructions for punctuation and formatting. AICHE infers these from your speech but doesn't take voice commands. If you prefer explicit verbal control, Willow's model fits.
Brand momentum and shipping cadence. YC X25 backing, $4.2M from BoxGroup, named angels (Dharmesh Shah, Alexis Ohanian, Max Mullen), and a Product Hunt 4.9/5 across nine reviews. Their team ships fast: Windows in January, Cursor support in February, Teams in March. Responsive support and an active Slack community get cited in independent reviews.
Optional offline mode (Mac and iOS). Local model fallback for connectivity-limited situations. AICHE is cloud-only on the transcription path. If you regularly dictate on a plane with no Wi-Fi and you're on a Mac or iPhone, Willow's offline mode beats AICHE's "queue it for later" behavior.
That's the honest list. The next section is longer.
Where AICHE wins
1. Platform coverage
AICHE runs on macOS, Windows, Linux, iPhone, iPad, Apple Watch, Android, plus a Chrome extension, an Obsidian plugin, and a public REST API. 9 platforms. Willow runs on macOS, Windows, iPhone, and Android. Four.
The gap isn't a rounding error. If you're a Linux developer (Ubuntu, Fedora, Arch, Pop!_OS, whatever), Willow doesn't exist for you. If you write inside Obsidian and want voice notes that land directly in your vault, Willow doesn't exist for you. If you want a Chrome extension that dictates into any web text field (ChatGPT, Gmail, Notion, Linear), Willow doesn't exist for you. If you want to record a thought from your Apple Watch on a walk without pulling your phone out, Willow doesn't exist for you.
For developers, the Linux gap is the loudest. Most modern engineering workflows include at least one Linux machine - a server, a WSL setup, a primary workstation. AICHE installs from .deb, .rpm, AppImage, or Flatpak and the global hotkey (Ctrl+Alt+R) works under both X11 and Wayland.
2. Real public REST API
AICHE's Pro tier exposes a callable REST API. You can write a script, a shell function, a CI step, an Alfred workflow, a Raycast extension, or a custom Neovim binding that hits AICHE programmatically and gets text back. Willow does not advertise a public API as of May 2026.
This matters specifically for the kind of person who reads comparison articles before buying a voice tool. If you'd want to wire dictation into your own automations, AICHE is the only option in the table.
3. Software Development profile
AICHE Pro includes a recognition mode tuned for code: API names, CLI flags, library names, snake_case and kebab-case identifiers, common dev jargon. You can dictate useEffect, --no-verify, kebab-case, argparse, kubectl get pods -n staging, and the transcription respects the formatting instead of "fixing" it into prose.
Willow's marketing emphasizes Cursor and VS Code integration. The IDE works fine. But there is no advertised code-tuned recognition profile on their pricing page or feature list. Dictating into Cursor still goes through the same general-purpose model that wants to write English.
If you spend hours a day talking to Claude Code, Codex, Cursor, or Antigravity, the recognition profile is the difference between "this saves me time" and "I keep fixing transcription mistakes."
4. Voice Code for AI coding agents (Pro)
AICHE Pro includes Voice Code: dictate straight into AI coding agents (Claude Code, Codex, Cursor, Antigravity) with pause-aware auto-send. Stop speaking for a beat and your prompt ships to the agent without an Enter press. Voice confirmations let you approve or reject agent-proposed actions by voice.
This is voice for the agent loop specifically, not generic web dictation. Willow's pitch is "voice replacing your keyboard" across all apps. Their Cursor integration is general-purpose dictation aimed at the IDE. AICHE's Voice Code is shaped for the back-and-forth of agentic coding.
5. Pricing
AICHE Personal: $3.99/mo (annual, $47.99/yr). 7-day free trial, no credit card.
AICHE Pro: $8.33/mo (annual, $99.99/yr). Same trial.
Willow Individual: $12/mo (annual, $144/yr). Free tier exists but caps at 2,000 words/week ("roughly 16 minutes of natural speech" per the Voibe review).
Over three years on the entry paid tier: AICHE Personal is $143.97. Willow Individual is $432. That's a $288 gap for what is in many respects the same product category. The same Voibe review notes Willow's "$144/year Individual matches Wispr Flow's headline rate" - it is not undercutting the field.
The free tier is a real difference too. AICHE gives you 7 days of the full product (no word cap, no feature gate). Willow gives you 2,000 words a week, indefinitely. A heavy user can exhaust that in one session - so the free tier is more "try it" than "use it."
6. Team plan accessibility
AICHE Pro includes team management (seats, roles, unified billing, shared admin) with no listed seat minimum. A pair, a solo founder with one assistant, or a five-person team all work.
Willow's Team tier is $10/mo per seat, billed annually, with a 3-seat minimum. If you're a solo with one EA, or a two-person founding team, or a small consultancy of two clinicians - you're paying for three seats whether you have three people or not. The Voibe review of Willow Voice (Voibe is an independent voice-app review site) calls this out directly: "3-seat Team minimum locks out 2-person teams - 2-person teams have to pay for an unused third seat ($120/year wasted)."
7. Compliance posture across paid tiers
AICHE's zero-retention audio model and E2EE sync are part of the product on both Personal and Pro - not gated to a custom-priced Enterprise SKU.
Willow advertises "SOC 2, HIPAA, zero data retention" on their homepage. Read the pricing page and those guarantees are listed under the Enterprise tier (custom pricing, billed annually), alongside SSO/SAML, MSA + DPA, and dedicated support. The Individual and Team tiers don't include the same contractual posture. If you're a solo therapist or a small clinic wanting a BAA at the Individual price point, that path doesn't exist - you go to Enterprise sales.
8. Specific numbers, not "fast"
AICHE publishes hard numbers:
- Sub-100ms cold start to recording (audio prewarm).
- ~3 seconds to transcribe 15 minutes of audio.
- within 1 second after processing server-side audio retention (typically 2-3s during processing).
- AES-256-GCM at rest for cloud sync, Argon2id key derivation, modern TLS, with certificate pinning on iOS and Android in transit.
- 99 transcription languages, identical engine on every platform.
Willow publishes a 200ms transcription-to-text latency on their own benchmark post. The Voibe review does not independently re-measure the millisecond figure and notes that "on strong wifi the experience is snappy; on weak wifi it lags." Their published privacy policy (last updated 2025-04-30) does not name a specific cloud transcription provider or LLM subprocessor and does not specify a server-side audio retention duration.
9. Language nuance worth knowing
AICHE handles voice input + transcription in 99 languages on every platform - same engine, same accuracy, same auto-translate-to-English option. A Russian speaker on Linux gets Russian transcription even though the menus are in English (UI localization is mobile-only; iPhone, iPad, Apple Watch, Android cover 28 UI languages).
Willow claims 100+ languages on their pricing page footer. Both products run on Whisper-class engines, so practical coverage is comparable; the marketing-count difference is mostly how each side counts. One Product Hunt reviewer (Martijn van Noordennen) flagged that "if you turn on fast dictation, it will sometimes translate the language in which you dictate to English, which... is not intended." If you dictate in multiple languages and want predictable behavior, that's worth checking against your own workflow before committing.
A note on privacy posture
Privacy on voice apps is one of those areas where the marketing surface and the contractual surface can drift. Here's the neutral way to read both apps.
AICHE's model, in plain text: Desktop audio streams to a named cloud transcription provider (Groq), gets processed, and is purged immediately after processing, within 1 second. Transcripts live locally on your device by default; cloud sync is opt-in and end-to-end encrypted with AES-256-GCM, key derived from your passphrase via Argon2id. We can't read your synced notes on our servers because we don't hold the key. Transmission uses modern TLS, with certificate pinning on iOS and Android. On desktop (macOS, Windows, Linux) there's no automatic telemetry - diagnostics flow only on user-initiated action (the OS-level "share with developer" prompt after a crash, or a manual "Send Diagnostic Report" button). On mobile, the only third-party telemetry is Firebase for ad attribution. There's no global keystroke logging, no clipboard monitoring outside explicit user action, no active-window-title transmission, no background recording.
Willow's model, as their public materials describe it: Cloud-first transcription by default. An optional Offline Mode on Mac and iOS runs a local model as fallback (not the default). Their published privacy explainer (and the privacy policy last updated 2025-04-30) distinguishes between an Opt-In mode that retains "recognized dictated text (anonymized)" and a Private mode that retains only operational telemetry (email, name, words-per-minute, word count, device info, performance metrics) without voice or dictated text. The policy doesn't name specific cloud transcription providers or LLM subprocessors. SOC 2 + HIPAA + zero-retention as a contractual package is gated to the Enterprise tier per the pricing page.
If you're comparing voice apps on privacy, here's what to look for in any of them, regardless of brand:
- Does the privacy policy name the cloud transcription provider?
- Is there a specific server-side audio retention duration in writing?
- Is "zero data retention" a contractual commitment on the tier you're actually buying, or only on the Enterprise one?
- Is the encryption-at-rest scheme published (algorithm, key derivation), or is "encrypted" a marketing word?
- Does the app log keystrokes outside of explicit recording? Monitor the clipboard? Transmit window titles?
These are reasonable questions to ask any voice app, including AICHE. The answers should be in writing somewhere you can link to.
Common questions
Q: I'm on Linux. Which one works for me?
A: AICHE. Willow doesn't list Linux support on their site (verified May 2026). AICHE ships .deb, .rpm, AppImage, and Flatpak with Ctrl+Alt+R as the global hotkey under X11 and Wayland.
Q: I dictate into Cursor and Claude Code all day. Which one handles snake_case and CLI flags better?
A: AICHE, if you turn on the Software Development profile (Pro tier). It's tuned for code identifiers, CLI flags like --no-verify, kebab-case, snake_case, library names, and shell syntax. Willow has Cursor and VS Code integration but doesn't advertise a code-tuned recognition profile, so dictation goes through the general-purpose model.
Q: I'm a solo founder. The team plan minimums matter to me.
A: AICHE Pro ($8.33/mo annual) includes team management with no seat minimum - one seat works, two seats work, ten seats work. Willow's Team tier requires a 3-seat minimum at $10/mo per seat annual. If you're solo, AICHE is the cheaper and cleaner path.
Q: I want a public API I can call from my own scripts.
A: AICHE Pro exposes a REST API. Willow does not advertise a public API as of May 2026.
Q: I need HIPAA for a small therapy practice. Two clinicians.
A: Neither app advertises HIPAA on the entry paid tier. Willow gates SOC 2 + HIPAA + zero-retention to the Enterprise tier (custom pricing, contact sales). AICHE doesn't market a HIPAA compliance package and ships zero-retention audio + E2EE sync across all paid plans, but you'd need to verify whether AICHE's existing posture meets your specific BAA requirements. For two clinicians needing an active BAA, both apps require a conversation with their sales/legal side.
Q: I want offline dictation on a plane.
A: Willow has an optional Offline Mode on Mac and iOS - that's a real advantage in your specific scenario. AICHE is cloud-first for transcription (you do get a crash-proof local queue that finishes processing when you're back online, but real-time on-flight transcription requires Wi-Fi).
Q: Style memory - does AICHE have it?
A: No. AICHE has AI cleanup (filler removal, punctuation, paragraph breaks) but does not learn a per-app voice. If "Slack-casual vs Gmail-professional automatically" is a feature you want, Willow has it and AICHE doesn't.
Q: What about Apple Watch?
A: AICHE records from Apple Watch (tap your wrist, talk, syncs to iPhone and every other AICHE app). Willow does not advertise Apple Watch support.
Q: Free tier - which is better for trying things out?
A: AICHE gives you 7 days of the full product, no word cap, no feature gates, no credit card. Willow gives you 2,000 words a week indefinitely - useful as a permanent floor, restrictive for a real trial (heavy users exhaust it in a session).
Result: Willow Voice is a polished Mac/Windows/iOS dictation app with a real "style memory" feature, an iOS keyboard, and a fast-shipping team. If you live entirely on Mac and iPhone, the dictation-to-style-per-app workflow is real and worth $144 a year. AICHE wins on platform coverage (9 platforms vs 4, plus API), public REST API, no-seat-minimum team plan, code-tuned recognition, and entry price ($3.99/mo vs $12/mo annual). If you're a developer, a Linux user, an Obsidian user, a solo or pair team, or anyone who'd want to call dictation from their own scripts, AICHE is the answer.
Try it now: Download AICHE (7-day free trial, no credit card) and dictate one prompt into whatever you're using to write today. If it feels like the right tool, you've found your voice layer. If it doesn't, you've lost 7 days and zero dollars.