This is a step by step guide shows how to install and set up Dante socks5 proxy server in Ubuntu 26.04 LTS.

Dante is a free open-source software allowing to route network traffic between clients and servers. With it, you may bypass network restrictions, e.g., visit network resources that’re not available via direct access.

You may also use it to hide your original IP address, so that the destination server only sees the Dante server’s IP. In my case, I use dante socks proxy to reduce latency for better gaming experiences.

Prerequisites:

To follow this tutorial, you need an Ubuntu 26.04 server with sudo privileges.

In my case, I have a VPS server with 1 core CPU, 1 GB memory, and 20 GB storage.

Step 1: Install Dante Server

Dante is available in Ubuntu and most other Linux Distributions’ system repositories. To install it, connect to your Ubuntu server, then run commands below one by one:

sudo apt update

sudo apt install dante-server

After installed the server package, it will automatically create a systemd service and run it silently in the background.

To check the service status, use command:

systemctl status danted.service

For Debian, run systemctl enable --now danted.service to enable and start the service if it’s not running.

If the service is enabled but failed to run, don’t worry, follow step 2 to edit the configuration file.

Step 2: Configure Dante Server

Dante by default loads /etc/danted.conf file for server configurations. The file by default has all configuration disabled, causing the service failed to run.

To edit the config file, you may probably firstly make a backup (or delete the file if you want) by running command:

sudo mv /etc/danted.conf /etc/danted.conf.backup

Next, run the command below to edit the file:

sudo nano /etc/danted.conf

The command will automatically create the file with empty content if you deleted the original file or used the mv command to backup.

When file opens, add the following content and edit accordingly.

logoutput: /var/log/danted.log
errorlog: /var/log/danted.errlog
user.privileged: root
user.unprivileged: nobody

# The listening network interface or address.
internal: 0.0.0.0 port=1080
internal: :: port = 1080

# The proxying network interface or address.
external: eth0

# socks-rules determine what is proxied through the external interface.
socksmethod: username

# client-rules determine who can connect to the internal interface.
clientmethod: none

client pass {
    from: 0.0.0.0/0 to: 0.0.0.0/0
    log: error connect disconnect
}

socks pass {
    from: 0.0.0.0/0 to: 0.0.0.0/0
    log: error connect disconnect
}

Here are more about the rules:

• logoutput: /var/log/danted.log – set the access log location (optional).
• errorlog: /var/log/danted.errlog – set the error log location (optional).
• user.privileged: root – run the background service as root, it will drop privilege afterwards.
• user.unprivileged: nobody – run normal proxy operation with user nobody.
• internal: 0.0.0.0 port=1080 – listen on all IPv4 interfaces on port 1080. If want, you may change the IP and port number accordingly.
• internal: :: port = 1080 – listen on IPv6 (optional).
• external: eth0 – replace eth0 with your server network interface name. Run ip route get 1.1.1.1 | awk '{print $5; exit}' command in server to tell.
• clientmethod: none – means any IP in the client pass {} can start TCP connection.
• socksmethod: username – you need to given username and password to authentication, even with clientmethod: none. Use socksmethod: none (though not secure) so any IP in socks pass {} section can access without authentication.
• client pass { from: 0.0.0.0/0 to: 0.0.0.0/0 ...} – the first gate (TCP connection) when client trying to access dante server. It means allowing access from any IP to any network interface in server. You may use from: 23.168.111.125/32 to: 0.0.0.0/0, so only certain IP (23.168.111.125 in the case is your PUBLIC IP) can access.
• socks pass { from: 0.0.0.0/0 to: 0.0.0.0/0 ...} – control what allowed clients may do. With socksmethod: username it needs password authentication. Also, you may replace IP accordingly.

The rules above are for TCP connections. You may add more client pass {} and socks pass {} sections with following rules:

• from: ::/0 to: ::/0 – allow accessing from any IPv6 address.
• command: udpreply – allow UDP associate.
• udp.portrange: 40000-45000 – set UDP port range.

So, it can be look like:

client pass {
  from: 0.0.0.0/0 to: 0.0.0.0/0
}
client pass {
  from: ::/0 to: ::/0
}

socks pass {
  from: 0.0.0.0/0 to: 0.0.0.0/0
  command: connect udpassociate
  log: connect disconnect ioop
}
socks pass {
  from: ::/0 to: ::/0
  command: connect udpassociate
  log: connect disconnect ioop
}

# Log UDP replies from remote hosts back to clients
socks pass {
  from: 0.0.0.0/0 to: 0.0.0.0/0
  command: udpreply
  log: connect disconnect ioop
}
socks pass {
  from: ::/0 to: ::/0
  command: udpreply
  log: connect disconnect ioop
}

After edited the file, press Ctrl+S to save, and Ctrl+X to exit.

Step 3: Create User for Authentication

For authentication purpose, you need to create a user in the server side by running the command below:

sudo useradd -M -s /usr/sbin/nologin USER_NAME_HERE

Then set a password for the user:

sudo passwd USER_NAME_HERE

Replace USER_NAME_HERE with whatever name that you want. Then, you may use the username and password to authenticate during the connection.

Step 4: Create Log Files and Change ownership

If you enabled logs, then you also need to edit the ownership of log files.

First, run command to create the log files:

touch /var/log/danted.log

touch /var/log/danted.errlog

Then, change the owner to noboby and nogroup:

chown nobody:nogroup /var/log/danted.log

chown nobody:nogroup /var/log/danted.errlog

Step 5: Configure Firewall

If you have firewall enabled in the Ubuntu server (run sudo ufw status), then run the commands below to open the ports for dante proxy server:

sudo ufw allow 1080/tcp

Here replace 1080 according to which port you use, and also run the command below if you have UDP association enabled:

sudo ufw allow 40000:45000/udp

Step 6: Apply Changes

Finally, restart the dante service to apply change:

sudo systemctl restart danted

And, do run the command below to check the service status:

sudo systemctl status danted --no-pager

Any mis-configuration in /etc/danted.conf may cause the service start failure.

Step 7: Try connecting to Dante server

After properly set up the server, you may try running the command below in a local Linux computer to start connection:

curl -v -x socks5://USER_NAME:PASSWORD@SERVER_IP:1080 http://www.google.com/

It should output something looks like:
*   Trying xx.xxx.xxx.xxx:1080...
* Host www.google.com:80 was resolved.
* IPv6: 2001:4860:482a:7700::, 2001:4860:4826:7700::, 2001:4860:482b:7700::, 2001:4860:4827:7700::, 2001:4860:4829:7700::, 2001:4860:482d:7700::, 2001:4860:4828:7700::, 2001:4860:482c:7700::
* IPv4: 142.251.153.119, 142.251.156.119, 142.251.150.119, 142.251.157.119, 142.251.152.119, 142.251.154.119, 142.251.151.119, 142.251.155.119
* Opened SOCKS connection from 192.168.0.104 port 49060 to www.google.com port 80 (via 23.106.152.194 port 1138)
* Established connection to 23.106.152.194 (23.106.152.194 port 1138) from 192.168.0.104 port 49060 
* using HTTP/1.x
...

For a graphical socks5 client, I use proxifier, which however is NOT a free software. You may try the free open-source ProxyBridge that works in Linux, Windows, and macOS.

Shotcut, the free open-source MLT based video editor, finally announced the 26.6 release after a month of Beta testing.

The new version of this Qt-based video editor introduced some new plugins support, HDR preview, Vulkan in FFmpeg, external monitor improvements, and various bug-fixes.

First, the new version introduced limited support for external audio and video plugins. Advanced users can now load OpenFX image effect plugins and/or VST2 and LV2 audio plugins, from the standard system directories.

By launching the video editor with new --experimental command line option, you’ll see a new “Manage Add-on Filters” setting gear icon in the add filter widget.

It will open a new dialog which lists ALL available filters, including the external plugins. And, you can select and de-select MLT filters to show or hide them in/from the main filter selection. This is useful for exploring filters, especially the host of OpenFx filters.

As the features above are experimental. The video editor may crash due to bugs. If Shotcut crashes within 30 seconds upon startup, it will restart in a safe mode without external (OpenFX, VST2) plugins. While, user may go to Settings -> Leave Safe Mode to re-enable them.

Besides experimental external filters support, Shotcut 26.6 also introduced Vulkan Display Method support for Linux. User can go to Settings -> Display Method menu to switch from OpenGL to Vulkan, though it needs an app restart to apply change.

In addition, the 26.6 version added back using an additional system display as an external monitor. A new Player -> External Monitor -> Preview Window (HDR) is added to open video preview in a new separated window in external monitor. It however does not work along with the “Use Old Video Output” option, and, the HDR preview is not working on Linux and Windows/ARM so far.

It as well added support for PQ HDR and the metadata dialogs to External Monitor -> DeckLink.

Other changes include:

• New Reduce Noise: Audio (RNNoise) filter.
• Support for HDR to Export.
• Add Dynamic range to custom Video Mode and Timeline Properties.
• New “Create Transitions on Overlap” option.
• Use native file chooser dialog for Linux Snap package.

Get Shotcut 26.6

The official release note, installer packages for Linux, Windows, macOS, as well as the source tarball are available in Github project page via the link below:

For Linux, select download the AppImage (for modern Intel/AMD), add executable permission from the Properties dialog, finally click Run to launch the video editor.

Tips: Ubuntu since 22.04 does NOT support AppImage out-of-the-box, open terminal (Ctrl+Alt+T) and run command to install the required library:

sudo apt install libfuse2

Ubuntu user may search & install the Snap package from App Center (or Ubuntu Software), while, an official Flatpak packageFlatpak package is also available for most Linux though runs in sandbox.

Ubuntu 26.10 Snapshot 2, the second monthly development release, was officially announced last night!

The new development release finally added back the .iso image for Intel/AMD platform, and, introduced new cdimage download tree.

The new snapshot release changed the cdimage.ubuntu.com download tree structure. All the release images are now located in /ubuntu/releases and /ubuntu-server/releases. And, the previous releases links still work through 301 redirect.

For Ubuntu releases that are being developed, the daily build images are now located in their own sub-directories, such as /ubuntu/noble/daily-*, ubuntu/stonking/daily-*, /ubuntu-server/noble/daily-*. See the discourse thread for more about the changes.

Known Changes in Ubuntu 26.10

Ubuntu 26.10, code-name Stonking Stingray, is planned for October 15. It’s short term release with 9-month support circle.

As far as I know, it targets Linux Kernel 7.2 and GNOME 51 desktop environment. And, it will probably feature an optional local AI integration, Speech to Text support, minimal secure boot Grub, and full Rust core utilities. The official release note page is ready but still empty at the moment.

Other changes according to the change archives include:

• KDE Plasma 6.7,
• LXQt Desktop 2.4.0,
• FFmpeg 8.1.1, Nginx 1.30.0, Nodejs 24.15, Lazarus 4.6.
• And many Lomiri (formerly Unity 8) and rust library updates.

Moreover, the missing iso image for amd64 CPU architecture type is added back in this snapshot release. Meaning you can now try it out in native Intel/AMD platform.

Get Ubuntu 26.10 Snapshot 2

NOTE: According to announcement, the snapshot 2 download link may not visiable for you due to the content-cache.

To download the .iso images for Desktop and Server, as well as Netboot tarball, WSL, and pre-install server images, go to the link below:

For the official flavors with other desktops or for education purpose, download them via the links below:

NOTE: this release does not include Ubuntu MATE 26.10 and Ubuntu Studio 26.10.

• EdUbuntu 26.10.
• KUbuntu 26.10.
• LUbuntu 26.10.
• Ubuntu Budgie 26.10.
• Ubuntu Unity 26.10.
• Ubuntu Cinnamon 26.10.
• UbuntuKylin 26.10.
• XUbuntu 26.10.

Future Ubuntu 26.10 Development Releases

The next snapshot 3 release is scheduled on July 30th. There will be also snapshot 4 releases in August. For more, see the table below.

NOTE: GNOME 51 and Kernel are developed by other teams, not Ubuntu. And, the kernel release dates are made according to this predictions.

Dolphin Emulator, the free open-source GameCube, Wii, and Triforce emulator, released new 2606 version today for Linux, Windows, macOS, and Android.

The new version of this popular video game console emulator finally implemented the Game Boy Player support, which was requested more than 16 years ago.

To use the Game Boy Player in Dolphin, simply load a GBA game via the new “Game Boy Player ROM” option in settings dialog, then run the Game Boy Player Start-Up disc or Game Boy Interface.

It will automatically attach the Game Boy Player hardware when necessary through INI settings. While, users can manually attach the emulated hardware by adding HSPDevice = 2 under [Core] in Dolphin.ini or a GameINI files.

Though you can now use Dolphin to play GBA games, such as Pokémon, on computer or Android, it’s NOT the best choice due to higher hardware requirements. User is still recommended to use a dedicated Game Boy emulators for better experience and more features.

Game Boy Player support in Dolphin should be considered a curiosity rather than the premier way to run these handheld games in an emulator. Not only does Dolphin have much higher system requirements, but dedicated Game Boy emulators will provide a superior experience in almost every way, with better fitting UIs and support for many more Game Boy features.

image from dolphin emulator website

Besides Game Boy Player emulation, the 2606 version also implement necessary hardware for The Key of Avalon. The game now is playable in Dolphin, but without a perfect experience.

Currently, entering the Segaboot menu will essentially softlock the game as it cannot reboot normally for unknown reasons. The game does not detect that prize cards have been dispensed, meaning there’s a message at the top of the screen after rounds telling you to take your prize cards. This message can be reset in the Segaboot menu. We hope to fix these problems in the future.

This version also introduced new Bloom Blurred graphics mod, which will render “fake HDR bloom” effects at any internal resolution without any issues. To use this mod, simply enable “Enable Graphics Modes”, then access the properties for a game and enable Bloom Blurred.

Other changes in the release include individual Wiimote audio mixer, allowing Wii Remote speaker audio to play over any additional speaker device, including controllers that expose an audio output!

It as well includes custom cropping support, so that you can crop the game screen whatever as you want. Though, the menus and other parts of the game will get cropped too.

Get Dolphin Emulator 2606

The new 2606 installer packages for Linux, Windows, macOS, and Android, are available to download in its website via the link below:

For Linux, select download the Flatpak, X86_64 for Intel/AMD or aarch64 for Snapdragon X or RasPi.

Then, install it either via your system package manager (if supported) or via command. Though, you probably need to enable Flatpak support first.

flatpak install drag-and-drop-flatpak-file-terminal

For Ubuntu user who prefer the native .deb packages, I’ve built the 2606 version into this unofficial PPA for Ubuntu 24.04 and Ubuntu 22.04.

To add PPA and install the package, open terminal (Ctrl+Alt+T) and run commands below one by one:

sudo add-apt-repository ppa:ubuntuhandbook1/dolphin-emu
sudo apt update
sudo apt install dolphin-emu qt6-qpa-plugins qt6-wayland

NOTE: The native .deb package (even Ubuntu’s official package) does NOT work in Ubuntu 26.04 and 25.10. It seems not building issue, but Qt6 compatibility as it outputs following logs at launch:

qt.core.qobject.connect: QObject::connect(QSocketNotifier, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QSocketNotifier, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QDBusServiceWatcher, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QTimer, QXcbConnection): signal not found
qt.core.qobject.connect: QObject::connect(QDBusListener, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QDBusListener, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QSocketNotifier, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QSocketNotifier, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QDBusServiceWatcher, QDBusConnectionPrivate): signal not found
qt.core.qobject.connect: QObject::connect(QIBusProxy, QDBusConnectionPrivate): signal not found
QIBusPlatformInputContext: invalid bus.
Segmentation fault (core dumped) dolphin-emu

Found the image viewers and editors, such as Loupe and GIMP, cannot load HEIF/HEIC images by default in Ubuntu 26.04? Here’s a quick guide show you why and how to fix.

Today I found that the default Loupe image viewer in Ubuntu 26.04 could not load my photo images imported from iPhone.

It said “Either the image file is corrupted or it contains unsupported elements“, and the more information page showed the error below:

emote error: org.gnome.glycin.Error.LoadingError: glycin-loaders/glycin-heif/src/main.rs:103:24: DecoderPluginError(Unspecified) Decoder plugin generated an error: Unspecified

stderr:
Setting process memory limit

Why HEIF/HEIC does not load in Ubuntu 26.04?

Someone has reported the issue to glycin project, the GNOME backend library for loading images. It has nothing to do with that library, but caused by missing of the libheif plugin.

According to the glycin source page, it uses the external libheif library for decoding AVIF and HEIF images. And, libheif has the HEIF/HEIC decoding support through either libde265 or ffmpegdec plugin.

HEIC/HEIF support is handled by libheif’s libde265 or ffmpegdec plugin

By running the command below in terminal, you’ll find that the glycin library in Ubuntu 26.04 was however built with libheif1 and only a few of its encoder plugins as dependencies.

apt-cache depends glycin-loaders

And, libheif1 package only depends on aomdec and dav1d plugins, recommends aomenc plugins, while all other plugins are marked as “Suggests” packages that are NOT required to be installed during the installation of libheif1.

apt-cache depends libheif1

In short, the core image loading library supports HEIF/HEIC decoding, but through the libheif library. The libheif in Ubuntu 26.04 has the HEIC decoder built as separated plugin, but NOT installed out-of-the-box.

Enable HEIF/HEIC Support

So, the solution is simply install the missing plugin, either libde265 or ffmpegdec, for the libheif library.

To do so, press Ctrl+Alt+T on keyboard to open terminal window, and run the command below to install the package:

sudo apt install libheif-plugin-ffmpegdec

Refresh cache with sudo apt update if required. And, here you may replace libheif-plugin-ffmpegdec with libheif-plugin-libde265 plugin for same purpose.

Also, install the libheif-plugins-all package for all libheif plugins, if you have apps that reply on the library for encoding HEIC or encoding/decoding VVC, AVC, HTJ2K and other supported file formats.

Reload Thumbnails

If you found the image thumbnails do not work for HEIF/HEIC in file manager, then try running the command below to clear cache:

rm -rf ~/.cache/thumbnails

Finally quit Nautilus (the file manager) via command nautilus -q to reload.

HEIC image and thumbnails load afterwards

GDM Settings, the settings app for GNOME’s login screen (aka GDM Display Manager), is officially paused!

The main developer and maintainer of the project, Mazhar Hussain, announced the project is on old in last week via this issue:

So, for now, the development of this app is on hold, and I’m not sure for how long. It could be few months, or it could be more than a years. I might not even come back to this project specifically.

However, I love Open Source, and I love GNOME. So, I will be back some time in the future for some of my open source projects (maybe/possibly even this app).

GDM Settings App

In case you never heard of GDM Settings, it’s a free open-source community maintained application that provides an easy to use GUI for configuring the login screen appearance and behavior in Ubuntu, Fedora, other Linux with GNOME Desktop.

With it, you can change the themes, fonts, background, accent color, and top-bar appearance in the login screen, and, configure the power, logo, welcome message, as well as mouse and touchpad behaviors.

In my case, I use the app in every Ubuntu desktops for custom login screen background image. And, some Linux Distributions, e.g., Ubuntu (24.04 only), ALT Linux, and OpenMandriva, include gdm-settings package in their official repositories.

Ubuntu 26.04 custom login screen adjusted by GDM Settings

Since the last 5.0 version, the work on the application has slowed down. Now some features, typically the “Apply current display settings” option, stop working in GNOME 50 due to the change of the default monitor configuration file location.

As the main developer explained, he was very tiring and burnt out due to a 9 to 6 job, and now recovering from the burn out. So, the development of the app is on hold!

GDM Settings is still working (the most functions) in Ubuntu:

Except the “Apply current display settings” that applies the resolution, refresh rate, etc settings from Gnome Control Center to login screen, most other functions still work in my case in Ubuntu from 22.04 to 26.04.

For Ubuntu 24.04, simply open terminal (Ctrl+Alt+T) and run the command below to install it from system repository:

sudo apt install gdm-settings libglib2.0-dev-bin

Then you may start the app from the Gnome overview or app grid.

For other Ubuntu releases, you may try the official Flatpak package, which may or may not work because it runs in sandbox and stuck at GNOME 49 runtime.

Or, use this unofficial PPA, by running the commands below one by one:

sudo add-apt-repository ppa:ubuntuhandbook1/gdm-settings
sudo apt update
sudo apt install gdm-settings libglib2.0-dev-bin

(Optional) To uninstall the deb package, use command:

sudo apt install --remove gdm-settings

Also run the command below to remove the PPA:

sudo add-apt-repository ppa:ubuntuhandbook1/gdm-settings --remove

Darktable, the popular free open-source photography and raw developer application, released new major 5.6.0 version today!

The new version introduced optional AI sub-system, UI/UX and performance improvements, as well as many fixes and other changes.

The AI sub-system is disabled by default. If you build Darktable from source, then you need to configure with -DUSE_AI=ON option. After that, you may go to AI preferences tab to enable and download AI models.

Darktable AI preferences

As you see in the screenshot above, you can one click download & install the following free and open source AI models:

• denoise NIND and raw denoise nind, to remove real‑world sensor noise from RAW photos with high detail retention using a neural network.
• mask sam2.1 hiera small, for mask generation (segmentation) on images and videos.
• mask segnext vitb-sax2 hq, high‑quality mask‑generation model.
• upscale realplksr, real‑world photo upscaling model designed to handle noise, blur, JPEG/WebP compression, and other real‑image artifacts.

It also supports loading local AI models in .dtmodel format. And, it has out-of-the-box support for using CPU hardware acceleration for AI inference, while GPU is also supported, though needs manual setup and external ONNX Runtime library in Linux.

After enabled AI feature and downloaded the models, you can then go to darkroom mask manager to trigger either SAM2.1 or SegNext AI model (only one of them can be enabled at the same time). Then, click on an object to generate a precise mask, shift+click to subtract, scroll or Ctrl+scroll to adjust the smoothing or opacity. And, it supports saving the mask as a PNG for external use.

And, go to neural restore module in left side-bar for AI-based raw denoise, denoise, and upscale. Default models are NIND UNet (denoise), RawNIND UtNet2 (raw
denoise), and RealPLKSR 2x/4x (super-resolution), and all runs on the ONNX backend. While, additional models are available in the model repository and can be
installed manually.

Besides the new AI features, this version also introduced new color harmonizer module with options to apply anchor hue, pull strength and width, protect neutral colors, and apply a Gaussian filter to smooth harsh transitions.

As well, it added support exporting to HEIF image format, lossy or lossless in 8/10/12 bit color depths with all supported color subsampling options.

Darktable 5.6.0 also features UI/UX improvements. It now has a welcome dialog allowing to set some configuration options at first launch. It now supports rendering and caching 6K or 8K thumbnail or full-screen preview images in lighttable view. And, the resolution of the darkroom view’s preview-resolution images has been increased from 720×450 to 1440×900.

It also added pinning images directly from the 2nd window, through either drag’n’drop from the the horizontal strip of thumbnails or keyboard shortcut, as well as touchpad gestures to darkroom and lighttable culling layouts, and smartphone-like simultaneous pinch zooming and two-finger panning for Linux and Windows.

Darktable Welcome dialog

Other changes include:

• New --library <path> command-line option.
• Include lua-scripts.
• Lua AI API (darktable.ai) for scripting AI model inference.
• New cameras support.
• Translation updates.

There are also numerous improvements, bug-fixes, and other changes not mentioned in this post. See the Github release page for details.

How to Install Darktable 5.6.0

Darktable provide official installer packages for Linux, Windows, and macOS, which are available to download via the link below:

For Linux on modern Intel/AMD platform, select download the AppImage, add executable permission from its properties dialog, finally run to launch the photography software. Though, Ubuntu since 22.04 needs to install libfuse2 package for AppImage support.

For Ubuntu user who prefer native .deb package, keep an eye on this OBS repository (not updated at the moment of writing), or use the unofficial PPA so far supports Ubuntu 22.04, Ubuntu 24.04, and Ubuntu 26.04.

To add the PPA and install Darktable, open terminal (Ctrl+Alt+T) and run the commands below one by one:

sudo add-apt-repository ppa:ubuntuhandbook1/darktable
sudo apt update
sudo apt install darktable

NOTE: If you want GPU accelerated AI inference, follow the official setup guide, then also install the ONNX Runtime library (for Ubuntu 26.04 only, 24.04/22.04 have it built in bundle) via command:

sudo apt install libonnxruntime1.23

After that, go to AI preferences setting page, choose GPU AI acceleration and find the ONNX Runtime library at:

• /usr/lib/x86_64(or aarch64)-linux-gnu/libonnxruntime.so.1.23 for Ubuntu 26.04.
• /usr/lib/x86_64(or aarch64)-linux-gnu/darktable/libonnxruntime.so for Ubuntu 24.04/22.04.

To uninstall the PPA and darktable deb package, use commands:

sudo add-apt-repository --remove ppa:ubuntuhandbook1/darktable
sudo apt update
sudo apt remove darktable

FFmpeg, the popular free open-source multi-media library, released the second maintenance update for its 8.1 release series a few days ago.

The new version of this media library comes with more than a hundred of changes, mainly include bug-fixes, stability improvements, as well as some minor new features.

First, the new version added two new option aliases gamma22 and gamma28 to FFmpeg’s global options table, so users can specify these gamma curves more conveniently when encoding or processing video.

It fixed potential crashes or undefined behaviors in the Snow video encoder where the function get_dc() could read past the edge of the image or the function get_block_rd() could call memcpy() with a size that was too large.

If also fixed JPEG 2000 decoder where the ROI (Region of Interest) shift‑up operation could trigger undefined behavior, memory‑handling bugs in SMPTE 436M -> EIA‑608 bitstream filter, a logic error in the x86‑optimized Vorbis DSP code, as well as a potential underflow bug in H.264 decoder.

It now enforces global pixel‑safety limit inside each JPEG tile, each JPEG tile of a CRI frame, and the H.264/HEVC decoders when they are invoked through IMM5, preventing oversized or malicious frames from causing excessive memory use.

And, it now rejects ProRes RAW frames whose tile alignment values are invalid, PSX ADPCM streams where block_align % channels != 0, AAC frames when the decoder ends up with no valid sample rate after decoding, and COOK streams where subpacket channel counts add up to more than the total number of channels, preventing out‑of‑bounds access and unsafe memory access.

For the libavformat demuxing and muxing library, it improved the SCTP (Stream Control Transmission Protocol) demuxer by adding a size validation check in sctp_read() that mirrors the one already present in sctp_write().

It now uses a safer, faster, and more robust reallocation method when growing the MPEG‑TS program list, sanitizes FTP paths and command strings to block CR/LF injection, Telnet control characters, and invalid characters in RNTO (rename‑to) operations, and, honors the user‑provided io_open callback in HLS, DASH encoding, and DASH decoding, enabling custom I/O behavior everywhere.

It as well introduced MOV‑related fixes by validating dimensions and tile sums, capping ICC profile copies, checking APV access‑unit lengths, preventing negative seek indices, and preserving audio layouts when chan/chnl parsing fails.

The showcwt filter now correctly fills EOF regions for DU and RL directions, avoids undefined float‑to‑int conversion when tracking consumed samples, and fixes an out‑of‑array read in compute_kernel, eliminating multiple correctness and safety bugs.

The drawtext filter now avoids multiple double‑free scenarios (in border‑glyph cleanup, aliased glyph freeing, and cached‑glyph handling), plugs several error‑path memory leaks, ensures HarfBuzz‑shaped text frees allocations on failure, and validates that glyphs use FT_PIXEL_MODE_MONO to prevent invalid rendering and memory misuse.

Other changes include:

• memory-overwrite fix when converting UYVY video to YUV422.
• fix for yuv2planeX function in the PowerPC (PPC).
• improve swscale supports feeding video frames in slices (chunks of rows) even when you are using multiple scaling steps chained together.
• add a proper validation check for the number of audio channels (nb_channels) in the misc4 decoder.
• fix NVENC encoder compatibility with Video Codec SDK 13.1.
• handle 1‑line MEDIAN slices correctly in MagicYUV decoder.

There are as well tons of other bug fixes and improvements. See the official changelog file for details.

Get FFmpeg 8.1.2

The source tarball for the new version is available to download in its website via the link below:

For Ubuntu, besides building from the source, you may try this unofficial PPA which so far supports Ubuntu 22.04, Ubuntu 24.04, or Ubuntu 26.04.

NOTE: The PPA packages are NOT well tested! It MAY break things. You must know what you’re going to do, and use it at your own risk!

Simply open terminal (Ctrl+Alt+T) and run commands below one by one to add PPA and install FFmpeg 8.1.2:

sudo add-apt-repository ppa:ubuntuhandbook1/ffmpeg8
sudo apt update
sudo apt install ffmpeg

(Optional) To uninstall, run the commands below one by one to install ppa-purge tool and use it to purge PPA and downgrade FFmpeg to the stock version.

sudo apt install ppa-purge
sudo ppa-purge ppa:ubuntuhandbook1/ffmpeg8

This tutorial shows how to install the latest version of fish shell, and set it as default in Ubuntu. Though the title says for Ubuntu 26.04, it also works on Ubuntu 24.04 and 22.04.

Fish, stands for Friendly Interactive SHell, is an extremely popular command line interpreter that reads your text input, interprets and tells OS what to run, and returns the output.

Compare to the default bash shell, Fish provides a faster and smoother user experience. It features smart auto-suggestion that displays predictions based on history and PATH, in a muted gray color directly after your cursor.

It has out-of-the-box syntax highlighting that colorizes commands, flags, variables, and errors as you type. As well, it features modern and clean syntax, and ships with thousands of completions for tools like git, docker, and kubectl.

How to Install The Latest Fish shell in Ubuntu:

Fish is available in Ubuntu repository, but always old!

For the latest version, it’s available to install through the official PPA (4.x series so far). And, so far it supports Ubuntu 22.04, Ubuntu 24.04, and Ubuntu 26.04 on AMD/Intel and ARM64 (e.g., Snapdragon X and RasPi) platforms.

1. First, press Ctrl+Alt+T on keyboard to open terminal, then run command to add the PPA:

sudo add-apt-repository ppa:fish-shell/release-4

For Ubuntu Server, you probably need to refresh cache and install the software-properties-common package first for being able to use the add repository command:

sudo apt update; sudo apt install software-properties-common

2. For Linux Mint, refresh cache, while Ubuntu does it automatically during adding PPA:

sudo apt update

3. Finally, install fish shell via command:

sudo apt install fish

The last command usually selects install the latest version. Just in case, you may run apt policy fish to check if the one from PPA is available, and no other versions with a higher priority (larger than 500).

Configure Fish Shell

Fish shell provides a web based UI, allowing to change its color scheme and prompt, as well as managing functions, history, variables, and bindings.

To start the config UI, simply run fish command to switch to the shell, and then run fish_config command to start the web server. It will automatically open the UI in your default web browser:

For Ubuntu server without a graphical desktop environment, then either edit config files under .config/fish (see the documentation for details) directory, or, connect to your server with port forwarding:

ssh -L 8000:localhost:8000 user@your-server

Finally run fish_config and open localhost:8000 in your local web browser.

Set Fish as Default Shell

After installed the package, you may run fish command every time to switch to the fish shell, and run exit to exit, or set it as default either for your terminal app only or login shell.

Set Fish as default for your Terminal app

For Ubuntu 26.04 with default Ptyxis terminal, simply open the terminal preferences dialog from its hamburger menu (≡). Then switch to Profile and click edit the current profile.

In the pop-up profile editing dialog, turn on “Use Custom Command” toggle under Shell section and input fish.

For Ubuntu 24.04/22.04, open terminal preferences and navigate to Profiles (Unnamed by default) -> Command -> Run a custom command instead of my shell, and input fish.

After that, every time you launch the terminal app, it automatically switches to the fish shell.

Set Fish as default for all command consoles

If you want to set fish as default shell for all terminals, tty console, SSH remote connection, then you may set it as login shell.

NOTE: Setting fish as your login shell may cause issues, such as an incorrect PATH!

1. First, open terminal and run command below to list all the valid shells in your system:

cat /etc/shells

If ‘fish’ is not in the list, run command -v fish | sudo tee -a /etc/shells command to add it.

2. Next, run the command below to change login shell to fish:

chsh -s "$(command -v fish)"

Note this only works for the current user that you execute the last command.

After that, you may re-launch terminal, switch to TTY console (Ctrl+Alt+F1~6), or re-start your SSH connection, it should automatically switch to fish shell.

(Optional) As mentioned, changing the login shell may cause issues. If something goes wrong, use the command below to switch back to bash shell:

chsh -s "$(command -v bash)"

NVIDIA 595, the latest production branch driver, received a new update yesterday for Linux users.

The new NVIDIA 595.84 release includes fixes for many games, sleep and wake up issue, and a few regressions introduced in the 580 driver series.

As you may know, the NVIDIA 595 series Linux driver improved Wayland and Linux gaming support by introduced VK_EXT_present_timing and VK_EXT_descriptor_heap support, as well as nvidia-drm.ko modeset=1 parameter enabled by default. It as well fixed GPU hangs and Xid errors in Black Myth: Wukong, an action role-playing game inspired by the classical Chinese novel Journey to the West.

The new 595.84 continues improving Linux gaming by fixed hangs or corruption problems in 007 First Light, a 2026 action-adventure video game based on James Bond franchise, and Assassin’s Creed Origins a 2017 action role-playing game. Though both of the games do not support Linux, they are playable through Steam Play.

It also fixed the black screens issue in the Total War: Warhammer III game, which is a regression introduced in the previous 595 driver releases.

Besides the fixes for the above 3 games, the new version also includes app hangs or black screen issues in the following games:

• Crimson Desert, a 2026 action-adventure game.
• Elden Ring, a 2022 dark‑fantasy open‑world RPG game.
• Elden Ring Nightreign, Elden Ring expansion released in 2025
• ExoDomia, a narrative-driven, action-adventure game set on the robotic world of Salakar V!
• Far Far West, a chaotic 1-4 player co-op shooter.
• Grounded 2, a survival game announced in 2025.
• Incursion Red River, a single-player/COOP extraction shooter game.
• John Carpenter’s Toxic Commando, a 2026 first-person shooter game.
• Paradise Nowhere, an Underrated Backrooms game.
• Screamer, a reboot of the 1995 racing video game with the same name.
• Star Rupture, a first-person open world base-building game.
• Windrose, a co-op survival game.

In addition to the fixes for the above games, the new driver also fixed resuming from sleep issue, when runtime D3 (RTD3) power management is enabled to allow GPU to enter the very low‑power state while the system itself stays fully on.

There are as well fixes for stuttering and reduced performance issue in some applications, due to CPU/GPU synchronization when multiple CPU threads waiting on the same Vulkan semaphore.

Other fixes include:

• blank screen issue for X11 apps using Vulkan Present extension.
• performance bug in OpenGL memory management.
• DKMS kernel module build failure.
• Missing mode timings and invalidated display mode issues introduced in 580 driver series.

Get NVIDIA 595.84

The official release note as well as the .run installer are available in NVIDIA website via the link below:

For average users, it’s however NOT recommended to install the driver through the .run installer, as it’s quite easy to break your system.

For Ubuntu, NVIDIA 595 driver has already been made into the official repository for Ubuntu 22.04 ~ 26.04, though it’s so far at version 595.71.05. See this step by step guide or Ubuntu documentation to install it.