Timg – Terminal Image Viewer
github.com
github.com
More details, including other methods, are described here:
Edit:
Heh, threw it at HN and it worked fine for the most part. Trying to search just resulted in a blank page though. And throwing it at some news pages didn't exactly go "well" (and i reminded myself why i have no love for The Verge). Funny thing is that without all the JS and CSS to parse, even the biggest news sites came up in a matter of seconds.
One more thing we can kiss bye bye with Wayland i guess.
I've been using Alacritty myself for a couple of weeks now. It's not too bad: the htop header bar won't render properly and catting random binary files crashes the terminal pretty quickly, but it's smooth and does everything a good "tmux VM" should do well.
I understand iterm2 on macOS is a significantly improved experience?
Note however that there exist effort towards moving the tty into user space. And one leading figure in that is currently involved with systemd.
So i suspect it will happen, it is just a question of when...
Their claim that only mosh is like this, still present on their WWW site, is not in fact true now.
I say "largely" because there are a couple of other nasty escape sequences, like (for example) OSC sequences to set window titles and terminal names, that binary files can also accidentally contain and that therefore have to be disabled. Removing ISO 2022 makes the problem far less likely to be hit, but doesn't quite eliminate it entirely.
https://github.com/daleroberts/bv
It uses GDAL as a backend.
Traceback (most recent call last):
File "./tv", line 82, in <module>
SAMPLING = {'nearest': gdal.GRIORA_NearestNeighbour,
AttributeError: 'module' object has no attribute 'GRIORA_NearestNeighbour'
timg built and ran as-is on Debian 8 with the libraries that it needs.
For example, it has a tendency to abuse boldface black on black as a way to get a dark grey colour, making it not work so well on terminals where boldface really is a switch to a boldface font rather than brightening the foreground colour. On such terminals, black on black text is invisible, boldfaced or no.
Sadly, it actually works internally in terms of a 4x16x16x16 ARGB colour system, and could work quite well with 88-colour or 256-colour terminals, and very well with terminals that have 256x256x256 RGB, were it not for the fact that most of the information in the internal colour representation is thrown away by the ncurses and CGA-like layers that are built on top of it.
Since we are all posting alternatives, here's another:
One problem that all of these programs share is something that you cannot really do much about absent user-supplied information, since it is something that isn't exposed in the abstraction of a POSIX General Terminal Device. It is probably more acute for tv than for the others, albeit that as I have mentioned I couldn't get tv to actually run without error to see what the effect is.
This problem is that there is a baked-in assumption about the glyphs of the characters. This is not the usual concern that glyphs are different shapes in different fonts. That's not really much of a concern for box drawing and block graphic characters.
The baked-in assumption is that glyphs are rectangular, with a 1:2 aspect ratio. This is the assumption that underpins the idea that U+2580 is two square pixels, one atop the other.
But this is not true even for 9x15 fonts, which yield a slight distortion with all of these programs, let alone for yet squarer fonts. Barack Obama looks very odd indeed when U+2580 is actually a 16x16 or 32x32 square, as it sometimes is. (http://fileformat.info/info/unicode/char/2580/fontsupport.ht... shows several square appearances, and it is square in one of the fonts that I use here.)
tv, with the way that it already works, is probably best placed to incorporate user-supplied information about the font's character cell aspect ratio. pxl could probably be extended fairly simply, too. The others including timg, I suspect would be harder.
Interestingly, M. Zeller has had an open issue about this for timg since 2016.
https://github.com/nabijaczleweli/termimage
A useful additional script: #!/bin/bash for i in `find $@` ; do termimage --size 150x33 $i ; done
An "ANSI" terminal need only support ECMA-48, which has 8 colours; as can be seen from the terminfo entry for the "ansi" terminal type. "true colour" is ISO/IEC 8613-6.
That said, one of the issues on timg's issues list deals in the sheer amount of superfluous data that are generated and that that thus have to be transmitted across an SSH connection or a serial device. There is no optimization of the colour change control sequences at all. pixterm similarly does zero optimization, but no-one has pointed that out yet.
* https://github.com/hzeller/timg/issues/13
pxl, mentioned elsewhere on this page, does optimize colour control sequences and indeed uses an ncurses-style double-buffer mechanism to optimize re-drawing.
pxl employs a far more complete third party library for its terminal rendering, named termbox. This has ncurses-style double-buffering for no-op redraw suppression and redundant colour control sequence optimization. It does not optimize as far as BSD SCREEN did in the 1980s, which made use of several more tricks to reduce the need for emitting lengthy cursor motion control sequences (such as carriage return, line feed, and the various relative-motion control sequences; and not optimizing away writing a character if it is shorter to just re-output the next cell position than to move the cursor).
The downside of termbox is that whilst it supports 8-colour and 256-colour terminal types, and reads the TERM environment variable and terminfo entry to determine colour capability, it does not support 256x256x256 RGB colour and hardwires the 8-colour and 256-colour control sequences to ECMA-48 and ISO 8613-6 indexed colour control sequences rather than employing the appropriate terminfo fields, even though it reads the terminfo database.
timg, pxl, and pixterm also only display 2 pixels per character cell, because they all use the same single Unicode character for rendering (making termbox, with all of its worrying about character widths and replacement characters, overkill to an extent). As mentioned by its author on this page, tv uses a whole bunch of the box drawing and block graphic Unicode characters for increasing the number of pixels per character cell.
But whilst tv does both 256-colour and 256x256x256 RGB colour, it does not do 8-colour or greyscale like pxl does, and it completely ignores the terminfo database and the TERM environment variable.