VoiceInk is a great open-source local dictation app. ThoughtMic is a different tool — a proprietary voice-to-vault pipeline that writes titled, tagged, linked Markdown into your Obsidian or Logseq vault. Two legitimate trades: open-source identity, or vault depth.
No FUD. No marketing cherry-picks. We respect VoiceInk and what open source means.
Pick VoiceInk if…
Source on GitHub matters to you. You want zero subscription pressure, the freedom to fork or build from source, and a vibrant community-developed Mac dictation app that drops text wherever your cursor is. You're not asking the dictation tool to also organize your knowledge vault.
Pick ThoughtMic if…
You live in Obsidian, Logseq, Foam, or a plain Markdown folder, and you want spoken thoughts to land as titled, auto-tagged, backlinked notes — plus a Friday review surface that actually clears your #inbox and a local MCP server so Claude Desktop can read your vault. You're paying for vault depth, not for dictation.
| VoiceInk | ThoughtMic | |
|---|---|---|
| Platform | macOS | macOS (iOS 6–8 weeks after launch) |
| License | Open source (GPL v3) · auditable | Proprietary · closed source |
| Pricing | Free (build from source) or one-time $25–$49 | $8/mo · $72/yr · $99 lifetime (200-slot cap) |
| Free tier | Free forever (build from source, no auto-updates) | Free forever up to 2,000 words/week |
| Transcription engine | Whisper / Parakeet (local) | Whisper Large V3 Turbo (local, Metal-accelerated) |
| Privacy | Local transcription · cloud BYOK for AI enhancement | Local-only voice + vault · ZDR cloud rephrase opt-in (text-only) |
| System-wide capture | ✓ Text at cursor anywhere | ✓ Text at cursor anywhere |
| AI enhancement | ✓ BYOK — Ollama, CLI, or cloud APIs | ✓ Built-in, zero-config (Pro) |
| Vault writing | — None | ✓ Native, vault-agnostic Markdown |
| Auto-titles | — Manual | ✓ Generated from content |
| Auto-tags & backlinks | — None | ✓ Suggested from your existing vault |
| Friday review surface | — None | ✓ Keyboard-driven two-pane review for #inbox notes |
| MCP / queryable from AI tools | — None | ✓ Local MCP — query your vault from Claude Desktop or Cursor |
| Best for | People who want auditable, free, open-source local dictation on Mac | Obsidian / Logseq / Markdown-vault users who want voice as structured input |
Open source is a real value. So is depth-of-product. These are different trades, and both are legitimate. Here's an honest take on what each tool does well — so you pick the one that fits your priorities, not the one with the louder marketing.
#inbox notes (⌥⇧ R). Discard, keep, promote, archive in five minutes.VoiceInk gets text out of your mouth and into your cursor. ThoughtMic gets thoughts out of your mouth and into your vault, with structure.
Both apps press a hotkey, run Whisper locally, and put text where you want it. That's where VoiceInk's job ends, and that's the job it was built to do well. ThoughtMic adds a second job: the same voice capture also writes a structured note into your knowledge vault — with a generated title, suggested tags pulled from your existing tag taxonomy, and backlinks resolved against notes you already have.
VoiceInk is GPL v3. ThoughtMic is proprietary. That's a real thing to weigh.
What you give up with proprietary: source-level audit, the right to fork, and a guarantee against the company disappearing. Those are not small things, and we're not going to pretend they are. The PKM-buyer who picks ThoughtMic is making a trade — paying for depth (vault enrichment, review surface, MCP, ongoing development) at the cost of source-level verifiability.
What we offer in partial answer: verifiable network behavior. Run Little Snitch. Watch the connections. Voice and vault never leave your machine. The only outbound call is the optional text-only rephrase to Groq Zero Data Retention, and you can disable it in one click. Every telemetry event we send is documented at /privacy/telemetry/ — aggregate counts, no content, one-click off. Closed source means you can't read the code. It doesn't mean you can't verify what the code does on the wire.
If you live in Obsidian or Logseq, you've already invested in a tag system, folder hierarchy, and a personal lexicon of [[wikilinks]]. A pure dictation app — even a brilliant one — treats every dictation as a fresh string of text. ThoughtMic treats every dictation as a candidate note in your vault, structured the way you've already structured the rest of it.
Capturing thoughts is the easy part. The reason most voice-note workflows fall apart is the Collector's Fallacy — you accumulate hundreds of #inbox entries you never re-read. ThoughtMic ships a dedicated weekly review surface: hit ⌥⇧ R, walk through every captured note, hit one of four keys (Discard / Keep / Promote / Archive). Most weeks, five minutes clears the inbox. VoiceInk has no opinion on what happens to your text after it lands; ThoughtMic does.
ThoughtMic ships a local Model Context Protocol server. Your vault becomes queryable from Claude Desktop, Cursor, and any other MCP-compatible AI tool — without leaving your machine, without uploading anything. VoiceInk's text never persists into a queryable structure; once it's in the cursor, it's the cursor's problem.
Both apps run local models. Transcription quality is in the same range — they share the same underlying Whisper architecture, and VoiceInk has added Parakeet V3 support which can be quicker for short utterances. End-to-end latency for both is in the 700–1500ms range on M-series Macs. If you want sub-300ms streaming, the only path right now is cloud-based dictation (Wispr Flow) — and you'd be giving up the local-first guarantee both ThoughtMic and VoiceInk hold to.
"VoiceInk is excellent for pure local dictation, especially if open-source identity is a hard requirement. ThoughtMic is a proprietary tool focused on a different layer: writing to your knowledge vault with enrichment. We're not trying to replace VoiceInk; we're solving a different problem."
No. ThoughtMic is a proprietary commercial app, and we're upfront about that. If open-source identity is a hard requirement for you, VoiceInk is the right pick — full stop. The trade we're asking you to make is depth (vault enrichment, review, MCP, ongoing roadmap) for source-level auditability. Whether that trade is worth it is your call.
We may open-source pieces over time — the Obsidian starter vault is already free, and the local MCP server is a candidate. We won't commit to anything we can't deliver.
VoiceInk transcribes voice to text at the cursor. That's a complete product on its own and a great answer to the dictation problem. ThoughtMic does that and writes a titled, auto-tagged, backlinked Markdown note into your vault, plus a weekly review surface and a local MCP server. If you don't keep a Markdown vault, ThoughtMic isn't worth paying for — use VoiceInk.
If you live in Obsidian or Logseq, the time saved on titling, tagging, linking, and reviewing is what you're paying for. The pricing is $8/mo, $72/yr, or $99 lifetime (first 50 at $49). Free up to 2,000 words/week to try it without commitment.
Fair question. Two ways to verify without reading our source:
(1) Run Little Snitch. Voice and vault stay local; the only outbound call is the optional text-only rephrase to Groq Zero Data Retention. You can confirm this on your own machine in a few minutes — no marketing claim required.
(2) Read the telemetry disclosure. Every event we send is documented at /privacy/telemetry/. Aggregate counts only, no content, one-click off. We don't bury anything in a privacy policy.
Closed source is a real downside. Verifiable network behavior is how we partly answer for it.
Yes. Different hotkeys, different output destinations, no conflict. A common setup: VoiceInk for general-purpose dictation into Slack, Mail, your IDE; ThoughtMic for capturing thoughts that belong in your knowledge vault. They solve different problems and they coexist fine on the same Mac.
ThoughtMic ships Whisper Large V3 Turbo running locally with Metal acceleration. VoiceInk historically uses whisper.cpp and has added Parakeet V3 support for quicker transcription on short utterances. Transcription quality is in the same range; both are batch-style (record-then-transcribe) with end-to-end latency in the 700–1500ms range on M-series Macs.
Neither is a streaming-fast experience — that requires sending audio to the cloud, which both apps avoid by design.
No commitment yet. The honest answer is that the business model is paid Pro, and the enrichment pipeline (auto-tag/title/backlink quality, MCP integration) is the moat. Open-sourcing the whole app would undercut the only thing keeping the lights on for a one-person operation.
We may open-source pieces — the starter vault already is, and the local MCP server is a candidate. Anything we promise here we'd promise in writing. For now, treat ThoughtMic as proprietary indefinitely.
We'll email you the day ThoughtMic is ready — with your download link and, for the first 50 signups, a $49 lifetime Founder's Deal code.
We'll email you at launch. No newsletter. Unsubscribe any time.
Works with Obsidian · Logseq · Foam · VS Code · any .md folder
More honest comparisons