boardcast

Talk, and the whiteboard draws itself.

Local speech recognition · Live hand-drawn boards · Meeting replay and export

Website · Download the desktop app · Changelog

English · 中文

boardcast turns a discussion into a growing whiteboard. Flows, notes, assignments, questions and rejected proposals are written, connected and crossed out as the conversation goes. Import a paper or a screenshot to discuss and mark up the source itself.

Speech recognition and the board model run on your computer by default. Once the models are downloaded, you can work offline. You can also connect an OpenAI-compatible API for a larger model.

The desktop app is a native rewrite in Rust on GPUI. The earlier Tauri app is no longer released, and a few of its features are not in the current app yet.

Get started · What you can do · Shortcuts · Models and privacy · Development

Get started

1. Download a package from Releases. They hold the same app and differ in the speech models inside:

   Package	Recognizes	Speech models
   …-macos.zip, …-linux.tar.gz, …-windows.zip	Japanese, English and Chinese, each with its best model	about 1.1 GB
   …-ja	Japanese, with ReazonSpeech	about 160 MB
   …-en	English, with Parakeet	about 660 MB
   …-macos-none.zip	Whatever your Mac's own recognizer supports	none

   Every package also carries the 334 MB model behind meeting search. Packages are built for macOS (Apple Silicon and Intel), Linux x64 and Windows x64.

2. macOS: the app is signed only ad hoc, so open it with Control-click › Open the first time. Linux: unpack the archive and run ./install.sh (into ~/.local, no root needed; --uninstall removes it and keeps your data). Later versions then install themselves. Windows: unpack the archive and run bin\boardcast.exe; there is no installer, and it does not update itself yet.

3. On first launch, Settings opens on Board model: download a model that runs on this computer, or set up a remote one. Speech recognition is already included.

4. Click Start listening (or press Space), or type a line in the transcript's input and press Enter.

An installed copy looks for a new version shortly after starting. It installs in the background and waits for you to restart, so a meeting is never interrupted.

What you can do

During the discussion	After the discussion
Talk and draw: flows, topics, nested items, questions, formulas and notes share one board.	Summarize: collect conclusions, decisions and to-dos into a summary you can edit, tick off and copy.
Circle and refer: circle things, then say "connect these two" or "star this", and the model knows what you mean.	Replay and fork: watch the board grow, seek, change speed, or start a new board from any recorded moment.
Correct the board: select something to strike, circle, underline, star or highlight it, change its words or remove it. Undo and redo cover all of it.	Trace who said what: hover a speaker to see their part of the board; jump from an item to the line behind it.
Tell speakers apart by voice and give them real names; the same voice keeps its name.	Find a meeting again in the sidebar, or search all of them, and carry on where it stopped.

Boards are drawn stroke by stroke in Chinese, Japanese and English. Light mode uses whiteboard markers; when the system is dark, the board becomes a chalkboard. The interface is in English or Chinese, following the system or your choice in Settings; board labels and summaries follow the meeting's language.

Circle and refer

Click Circle in the tool bar under the board, then drag around things on the board. Draw as many loops as you like; everything with its middle inside a loop is taken in, up to eight. The next line, spoken or typed, carries those targets to the model, and the circles go away. Esc or the same button cancels. Circles never become part of the board or its exports.

Discuss your source material

Use Import in the sidebar, or drop a file on the board:

Source	What you can do
PDF / papers	One board per page with the page underneath. Mark words, sentences and paragraphs; link regions and add notes.
arXiv	Paste a link or paper ID to download the PDF.
PNG / JPEG / WebP / GIF	Read the text in a screenshot or a photographed whiteboard with the local vision add-on.

Take the meeting with you

Export	Use it for
Markdown	Key points, a Mermaid flow chart and the transcript.
PNG / SVG	This board, as it looks on screen; SVG stays sharp at any size.
Animated SVG	This board, written out stroke by stroke in the order of the meeting.
GIF	The same, for chats and documents that don't play SVG.
Video (WebM)	The same as an AV1 video: smaller and sharper than the GIF, for browsers, GitHub and web pages.
Replay page	One .html file for someone who missed the meeting. Nothing to install.
Whole meeting PDF	A page per board, then the summary and transcript, with selectable text.

Exports always use marker on white paper, whatever the window's theme.

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗