Friday, October 30, 2026

The Future of Human Interface for Emacs

When I'm "on the road," Vakana1 becomes my human interface to Emacs.

I use Vakana because its presence on my Android device affords me these three conveniences:

  • Portability
  • Speech recognition
  • Export options that include:
    • Org
    • Markdown
    • HTML
    • PDF
    • Slides

My computer is portable, and it does have a microphone. But its ability to recognize speech is poor despite multiple training sessions.2 As a workaround, I'd dictate an email message on my cellphone, save it as a draft, and then access it from my computer.

As speech interfaces, my Android email clients are clunky. But the real issue with them is that they distract me with unread messages. It's like working in the office on a Monday morning where chatty coworkers talk about how well their team played, or the latest attempt at a ceasefire, or the ever-increasing price of … everything. I needed a tool devoted to ideation. So I tried Vakana.

After a one-time setup with a "blog notes" thread, I simply create a new note and start dictating right away. It's like working late on a Friday evening with only unobtrusive air handlers as coworkers.

Dictation isn't perfect -- I resist the urge to edit except in cases where I misspeak and want to choose a different word or phrase. I focus primarily on capturing ideas; creating content is secondary.

When I'm ready, I'll export either to HTML or Org. Either way, all the thread's notes are combined into a single file. The HTML export produces a single HTML file with bead images included with an SVG tag. The Org export creates a ZIP file with several Org files, one for each note, plus SVG files. Then I'll attach the file to an email message, which I save as draft and open on my computer.

With the exception of Org export, this has nothing to do with Emacs.

When I wonder how I might operate Emacs with voice, I vaguely envision "voice-bindings" -- a software layer that converts a spoken command such as "Visit File" into its associated function. The subsequent input at the Find file prompt would be managed with a voice-controlled dired package, perhaps called diredv. And, by-the-way, it doesn't need to be a monotonic voice; it can be a melody where pitch might fill in for the value of a prefix argument. Rhythm might denote yet another aspect of a command or input. I laugh when I think of what might happen to someone with hiccups.3

Pursuit of this is not on my agenda.

I imagine that evolution of enhanced Emacs input would be driven by our community's most brilliant contributors. I'm thinking of the ones who use elaborate hardware devices tailored to their anatomy and workflow. Their devices resemble split keyboards. Sometimes modifier keys are positioned under the thumbs to minimize pinkie strain. You might've seen examples if you read r/emacs.4 The evolution from that might be gloves used in conjunction with a virtual keyboard.

Personally, I distrust specialized hardware. I've had many peripherals become obsolete due to the lack of an OS-compatible driver. As well, I don't over-customize Emacs. I think it's important that I can use an Emacs -Q invocation if necessary.

I'm sure that the FOSS community would ensure that drivers for a specialized Emacs input device would be free, even if only as source code. However, the real impediment would be getting employers to allow such specialized and low-level code to be installed on their systems. And I can't imagine anyone investing in a solution that can't be used for most of the day.


1 Learn about Vakana here.

2 The speech recognition training probably requires a registered, logged in account, which I refuse to create.

3 This reminds me of the motion-activated radio on board the Heart of Gold. You don't need to get up to change the channels or mute it. On the other hand, you have to sit awfully still if you don't want to change it inadvertently. Obviously the engineer who came up with that idea wasn't too fond of dancing.

4 The Emacs subreddit can be found here: https://www.reddit.com/r/emacs/

• This is my contribution to the October 2026 Emacs Carnival.

• One might jokingly refer to "The Future of Human Interface for Emacs" with the acronym "FUHIE," self-deprecatingly pronounced "Phooey."

Thursday, October 29, 2026

Highlight Active Emacs Window

In https://emacs.stackexchange.com/a/36240/47342, CodyChan provides a few options "to better distinguish between focused window." But the one I like best is highlight-selected-window, which, rather than highlighting the selected window, changes the background color of all the unselected windows. The effect is that the active window stands out in comparison to the others.

The especially neat thing about this solution is that when the minibuffer is active, the background of all the windows is altered. So it's obvious even when Emacs is expecting input -- something I often miss.

I've changed the color of the buffer-face-set background from "#111" (sienna4) to "antiquewhite2" because I still need to read the content of inactive windows. The original color is too dark.

Demonstration of highlight-selected-window

I've also adopted flash-active-buffer, which sets the background color of the active window temporarily. I changed the color to palegreen just because. I use s-? as a keybinding. It's a bit awkward, but it's as if I use "?" to ask, "Where am I?" The function highlight-selected-window works so well that I don't think I'll need flash-active-buffer.

The color names that Emacs understands are stored in color-name-rgb-alist.

See also https://www.color-hex.com/color-names.html, but note that Emacs does not support every color name listed here.

Saturday, October 24, 2026

A Point-Based Tooltip for Emacs

When you hover the mouse pointer over a link in Org, Emacs displays a tooltip that shows the link's URL.0 This mimics the behavior of a web browser that displays the URL in the status bar.

Within Emacs, I try to avoid the mouse as much as possible. I already configured Emacs so that pressing ENTER on a link follows the link.1 But this method doesn't provide a way to preview the link before following it. I wondered, "What if Emacs displayed a tooltip whenever point is on a link?"

The answer came to me immediately; "It would be darn annoying."

So I decided that the user should invoke a function to "display tooltip at point." It didn't take long for me to find a solution on Stack Exchange, "farynaio/org-link-copy."2 The function's primary purpose is to add the URL to the kill-ring. But it also shows the URL in a message that indicates the URL was copied. It works perfectly.

However, the tricky part was binding a key to the function. It was suggested that one could bind C-x C-l3 to the function with this statement:

(define-key org-mode-map (kbd "C-x C-l") 'farynaio/org-link-copy)

Unfortunately, when I included it in my init file, it generated the error "Symbol's value as variable is void: org-mode-map" even while it worked after Org was enabled.

Eventually, after trying (unsuccessfully) to add the assignment to org-mode-hook, I found it easy to include this statement above the define-key statement: (require 'org-keys). This provides a definition of org-mode-map, which is provided in org-keys.el.

As Emacs is very keyboard-oriented, this little addition allowed my fingers (and thumbs) to stay on the keys.

Recall that my original objective was to display a tooltip, not a message. However, a tooltip would appear at the location of the mouse pointer, not at point, even if the mouse is hovering over another program's window. Just try it yourself:

(tooltip-show-help "MSG")

But this disadvantageous behavior of tooltip-show-help could be useful if you wanted to locate the mouse pointer or signal the status of some long-running asynchronous Emacs process. Setting the message to "****\n****\n****\n****\n****\n****\n****\n****" might make the mouse obvious enough to notice. Feel free to play with that!

Trying to locate the mouse pointer leads me to another issue: how to locate point. So far, I've been using my overwrite-mode indicator to show me point.4 But I hope to devise a more effective solution soon. Perhaps.


0 If you don't see this, you might need to enable tooltip-mode.

1 Setting the variable org-return-follows-link to non-nil achieves this.

2 See https://emacs.stackexchange.com/a/60555. The function is org-link-copy.

3 I chose to bind the function to "s-SPC".

4 See Change Emacs Cursor to Indicate Overwrite Mode

Thursday, October 1, 2026

What is spook?

What is spook?

The Emacs function spook is not what you think it is. It doesn't make creepy noises, draw scary ASCII art or change the theme to orange on black. In fact it has nothing at all to do with Halloween, ghosts, monsters, pandemics or IRS audits.

What it does is insert words and phrases into the current buffer "to help give your message ... attractiveness for automated keyword scanners."

What kind of words or phrases might you expect to see? To find out I invoked spook while composing this content. Here's what it added:

TELINT Drug North Korea Security Consulting Virii Abu Ghraib Fort Hancock CUD AMEMB Vaccine MARTA racal E.O.D. Blizzard Nigeria

It's easy1 to configure spook to insert an alternate set of words and phrases -- just change the file name in the variable spook-phrases-file from "spook.lines" to the name of the alternate file. Unfortunately it must reside in the Emacs data directory; the file's location is fixed. The variable can be found in the Spook customization group, whose parent group is Games.

Even with only 72 lines in its source code file,2 spook demonstrates how to insert one or more random lines from a file into the current buffer.


1 It's "easy" if you have Administrator permissions.

2 The 72 lines is the entire spook.el file; the first 48% of the file provides the usual comments that describe the GNU GPL, author, maintainer info as well as documentation.