Short answer: open a Heptabase whiteboard or card, or annotate a PDF highlight, place the cursor in the block editor, press ⌃+⌥+R (Mac) or Ctrl+Alt+R (Windows/Linux), dictate the card body or highlight summary, press again. AICHE inserts text. You drag cards, add tags, and link backlinks on the canvas.
Heptabase objects: whiteboards, cards, sections, tags, PDF highlights, mind maps. AICHE writes card text; you organize spatially.
The Problem
Heptabase rewards thinking in cards. A whiteboard with twenty well-written cards beats one bloated document. But the friction shows up the moment you sit down to actually write the cards: every idea has to survive the trip from "I can see how this connects" to "I am typing a paragraph into a small card editor."
So most cards end up as one-line titles. The real reasoning lives in your head, or in the journal as a wall of half-sentences, or never gets written down at all. The canvas looks structured. The cards are thin.
What Changes
You stop writing cards like they cost something. Open a new card, hit the hotkey, talk through what you actually mean for two minutes, hit it again. AICHE drops a clean paragraph or two into the block editor. Then you drag it onto the whiteboard and connect it to the cards that surround it.
Math: speaking lands around 150 WPM, typing around 40. A 200-word card that took five minutes to type takes about 80 seconds to speak. The cards get longer because the cards are no longer expensive.
How It Works
- Open Heptabase. Anywhere with a block editor works: a card, a journal entry, a PDF annotation, the Inbox.
- Click into the block where the text should appear.
- Press ⌃+⌥+R (Mac) or Ctrl+Alt+R (Windows/Linux).
- Speak the full thought. Treat the card like you are explaining it to a colleague, not labelling a box on a diagram.
- Press the hotkey again. AICHE transcribes, cleans up filler, inserts clean text at the cursor.
- Drag the card onto the whiteboard, or
@-mention it from a journal entry, or tag it.
The flow is the same in the desktop app on macOS, Windows, and Linux. AICHE is desktop-only and inserts into whichever editor has focus, so the Heptabase native app and the web app both work.
Journal: Capture Before Organizing
The Journal is Heptabase's pressure valve. You are supposed to dump fleeting ideas into the daily entry and structure them later by dragging selections onto whiteboards or @-mentioning existing cards.
Voice fits this exactly. Open today's journal, hit the hotkey, narrate what just happened in your head: the meeting takeaway, the article you half-read, the connection you noticed between two cards you saw yesterday. You get a usable paragraph instead of a four-word stub you have to expand later.
When you come back to structure the day, the entries are detailed enough to drag onto a whiteboard as a real card. Heptabase converts the dragged selection into a note card automatically, and the card already has body content because you actually said something.
Cards: Reasoning, Not Labels
The cards that hold a whiteboard together are usually the explanatory ones - the "why this connects to that" cards, the synthesis cards, the trade-off cards. Those are the cards that suffer most from typing fatigue, because they are the longest ones and the ones you are least willing to start.
Click into a new card, hit the hotkey, talk through the argument. Constraints, edge cases, the source you are pulling from, the card you are reacting to. Stop, insert, edit lightly. The card now has the reasoning, not just a title.
This is also where AICHE's optional text cleanup earns its keep. Filler is removed, sentence breaks are added, lists become lists. Heptabase's block editor accepts the result without reformatting, so headings, bullets, and paragraphs land in the right blocks.
Mindmaps And Sub-Whiteboards
Heptabase's mindmap nodes are cards in a different layout. They behave the same way: each node has a title and a body, and the body is where the thinking goes. The fold control (Cmd + Shift + Enter on Mac, Alt + Shift + Enter on Windows) is more useful when the nodes contain enough content to be worth folding in the first place.
Sub-whiteboards have the same problem. A sub-whiteboard is a card that opens into another canvas, which means the cover card needs a description that tells you what is inside. Dictating that description is faster than typing it, and you tend to write a better one because you are explaining the sub-board out loud instead of compressing it into a label.
PDF And Web Card Annotations
Highlight a passage in a PDF card or a Web Card, create a highlight card from it, then dictate the note that goes next to the highlight. This is the workflow that gets skipped under typing pressure - you highlight everything and annotate nothing.
With voice, the annotation costs about the same as the highlight. You say what the passage meant to you, why you saved it, and which whiteboard it belongs on. The highlight card carries actual commentary onto the canvas.
Multilingual Capture
Heptabase has a heavy non-English user base, particularly in Taiwan where the company is based. AICHE handles multilingual input directly: speak Mandarin, Japanese, German, Spanish, whatever your thinking language is, and AICHE transcribes in that language. Turn on auto-translation in AICHE settings if you want to keep your cards in English while thinking in another language.
This matters more in Heptabase than in a chat tool because the cards persist. You are not throwing away a prompt after one use - you are building a knowledge base that you will reread for years. Capturing the original language often preserves nuance that translation flattens.
What You Get
- Unlimited voice notes with AI cleanup - filler removed, punctuation and paragraph breaks added, blocks land cleanly in the editor.
- Custom vocabulary - drop in the names of your projects, sources, authors, and recurring topics. They spell correctly every time.
- Multilingual input plus auto-translation - capture in your native language, optionally output English.
- Offline queueing - dictate on a plane, AICHE processes the queue when you reconnect. Heptabase is already offline-first, so the pairing matches.
- Zero-retention audio - audio is discarded immediately after processing, within 1 second.
Plans start at $3.99/mo (annual) with a 7-day free trial, no credit card. See pricing.
Common Questions
Q: Does it work in the Heptabase desktop app, the web app, or both?
A: Both. AICHE inserts into whichever text field has focus on the desktop. The Heptabase desktop app (macOS, Windows, Linux) and the web app in a browser both qualify. AICHE itself does not run on mobile, so the iOS and Android Heptabase apps are out of scope.
Q: Does it preserve Heptabase's block formatting?
A: AICHE inserts text where the cursor is. If the cursor is in a heading block, it lands as a heading. If it is in a bullet list, it lands as bullets. AI cleanup adds paragraph breaks, which Heptabase reads as new blocks, so longer dictations split into multiple paragraph blocks naturally.
Q: Can I dictate directly into a card title?
A: Yes. Click into the title, press the hotkey, say the title, press again. For card bodies, click into the body block first. Titles work best with short dictations - one sentence or less - because Heptabase wraps long titles awkwardly.
Q: What about @-mentions for backlinks?
A: Dictate the surrounding sentence, then type the @ and pick the card from Heptabase's picker. Voice is for the prose around the link, not the link itself. The mention UI is faster with the keyboard.
Q: How does this interact with the Heptabase CLI and AI Agent?
A: They are different layers. The CLI and AI Agent work on cards programmatically. AICHE is the human input layer - it gets your words into the editor. You can use AICHE to dictate a prompt for the AI Agent in the same way you would dictate it for Claude Code or Cursor.
Q: My cards mix English and technical terms. Will it mangle the technical words?
A: Turn on the Software Development profile (Pro) if you write a lot of code or product terminology. Add specific terms (project names, libraries, internal jargon) to your custom vocabulary and they come back correctly spelled.
Result: the cards on your whiteboards stop being thin labels and start being the real thinking. Capture costs less, so you capture more, and the canvas reflects what you actually know instead of what you had energy to type.
Try it now: open Heptabase, create one card on a whiteboard you have been neglecting, press your hotkey, and spend ninety seconds explaining why that whiteboard matters. Drop the card in the middle of the canvas and let it anchor the rest.