Shotcut 26.11.15 was released on November 15, and its own notes call it mostly a release that repairs problems from the last several versions, with one new tool added: an Audio Mixing Assistant. The fixes are listed one by one and are easy to check against what broke in 26.9, but the assistant is documented mainly in a pull request that its author called a proof of concept, so there is less on record for it than its name suggests.

What the release fixes

The Shotcut 26.11.15 release notes on the project blog list nine regressions from 26.9 as fixed. They include the crash when drag-and-drop offered to convert a video while Automatically Add Tracks was on, scratchy audio playback, broken audio scrubbing in the Source player, and a crash when hardware decoding was combined with preview scaling in GPU processing mode. Copying several clips that include dissolve transitions works again, and so do pasting filters onto a track and drag-and-drop clip moves, which had become slow. Two of the nine affect only Windows. Three older regressions are also closed: VST2 plugins hanging at startup or crashing on exit (from 26.8), and, from 26.4, Typewriter filter timing in preview and Glaxnimate failing to launch from Program Files on Windows. For a release that is mostly repair work, compare LibreOffice’s 26.8.3 point release with its 45 bug fixes.

Shotcut 26.11 Released With A Rule-Based Audio Mixing Assistant And Fixes For 26.9 Regressions

Linux packages and performance

The fix aimed squarely at Linux users is a file-dialog crash in the AppImage and Snap builds on Ubuntu 26.04.1. The notes don’t say whether Flatpak or distribution packages were affected. Other fixes cover playhead and marker snapping when the Timeline is zoomed out, and Subtitle Burn In stopping at items with blank lines. Lower memory use for audio transitions and faster dissolves in 8-bit processing mode are claimed without figures or a test method, so for now they are statements, not measurements.

What is new since 26.9

Proxy detection now follows Premiere and Resolve naming: a Proxy or Proxies subfolder, in any letter case, with an optional _Proxy suffix on the file name. That helps anyone who moves projects between editors. GPS filters can now read plain-text logs, such as dashcam output, and load a GPS file with the same name automatically. Elements has a search field that works only in English, the playhead is thinner, and new settings let you turn off the rounded rectangles that came with the 26.9 redesign or sync zoom with keyframes. The add-on filter UI generator gains rectangle parameters but is still labelled experimental. Anyone upgrading straight from 26.9.6 or earlier also picks up the 26.9.27 security fixes. Those followed a high-severity advisory on the update check falling back to plain HTTP, although the advisory still lists no patched version.

The Audio Mixing Assistant

The assistant sits under Timeline > Edit, and the release notes give it one line. The details are in pull request #1905, which proposed the Audio Mixing Assistant on August 29. You give each track a role (dialogue, background music, ambience, sound effects or a premixed program) and set a target loudness. The tool then scans the audio and suggests filters for each role. Dialogue gets gain moved toward the target, low-frequency noise removal and leveling. Music is set below the other audio, drops under dialogue and gets a mid-range cut. You choose which suggestions to apply, and they go in as one undo step, tagged so you can see later where they came from. Running the tool again replaces its earlier filters, so they don’t pile up. Ducking is done with keyframes on a gain filter, which builds on the automatic ducking added in 26.9.

“This tool is based on simple rules and creates static filters (some use keyframes for ducking).” — Brian Matherly, Shotcut developer, in pull request #1905

The record is untidy. GitHub still shows that pull request as open, and the only reply from lead developer Dan Dennedy says he wouldn’t get to it for a while. The project’s GitHub releases page also lists only the v26.11 beta (tag v26.10.30, October 31), not 26.11.15. The shipped code may therefore differ from what the pull request describes.

A loudness target the final mix isn’t held to

The assistant asks for a target loudness, but by its author’s own account it makes one pass over the audio and then writes fixed filters. The pull request lists two things as ideas that weren’t built: a loudness filter on the main output that would guarantee the final level, possibly with a second pass, and per-clip normalization for dialogue recorded at uneven levels. Unless the shipped version goes further, edits made after the run are not tracked until you run it again, and nothing checks the export against the target. Anyone delivering to a platform’s loudness limit should still measure the rendered file with Shotcut’s loudness meter. The assistant suggests changes and applies them, but it doesn’t check the result.

The notes give no date for the next release or for the unbuilt loudness pieces.

Christopher is a Linux and network administrator who writes about server management, security, and networking for Command Linux. Over his career he has managed web servers, mail servers, and small business networks, and his tutorials reflect problems he has solved firsthand. He is especially interested in server hardening and helping readers secure their systems from day one.