Our WolfCut source build stopped after 823 seconds because pkg-config could not find libavutil. If you want to edit video, that tells you to separate the contributor setup from the packaged application. If you want to change the code, FFmpeg library discovery is your first checkpoint.
The public WolfCut repository now redirects to Concat, a beta video editor. Its current README describes multi-track cutting, local captions, speech generation and background removal. Those features make it worth investigating for creators who want local editing without an account. Our recorded run never reached the editor, so it cannot settle whether Concat handles your footage well.
What happened when we ran it
Our sandbox installed WolfCut's dependencies in 52 seconds, recording 658 dependencies, then failed the build with exit 101 after 823 seconds. The test command also failed with exit 101 after 23 seconds. We ran commit 60ecb4a on October 17, 2026. These are source-workflow results, not editing or export benchmarks.
Both failing tails identify the same obstacle: ffmpeg-sys-the-third required libavutil, and pkg-config could not find its metadata. The diagnostic names libavutil.pc and says its parent directory must be in the search path. PKG_CONFIG_PATH was unset. That supports a precise conclusion: the required library was not discoverable through this check. It does not distinguish absent development files from files installed somewhere the lookup did not search.
The recorded configuration was 3 CPUs, 12 GB RAM and image lab-rust:1. That image label does not establish the compiler or operating-system version. The project lived under src/ in a 33.6 MB checkout. The scan found six CI workflow files and no separate tests directory. Rust checks can live beside source files, so the directory observation says nothing about assertion coverage.
There is no passing-test count or rendered video in this record. The failed test command encountered the same prerequisite as the build. Calling this an editor failure would overstate the evidence, and promising a successful build after fixing the lookup would do the same.
Source contributors need FFmpeg development files
The current README requires Rust 1.93 or newer, FFmpeg 7+ development libraries, cmake and a C++ compiler. It starts the source workflow inside src/. These are current upstream requirements, which should be checked against a pinned checkout before reproducing our October run.
The distinction between an FFmpeg executable and discoverable development metadata matters here. Our log specifically asks pkg-config for libavutil. Before repeating a long compile, check whether the intended environment can locate the documented native libraries. Record any later attempt separately from this blocked run.
For creators, the README points to downloads on the website and GitHub Releases. We did not evaluate those packages. A packaged beta may be a more appropriate trial route than compiling, but its existence supplies no evidence about installation ease on your machine.
Local captions require a model download first
The README says caption, voice and cutout models download through Settings and work offline afterward. That is a useful distinction for an offline workstation: obtaining the app and preparing its optional models are separate tasks. The project describes local Whisper captions and voice cloning, but our blocked source run measured neither accuracy nor output quality.
Use copied footage for the first project. Include the captions or voice work you actually need, then save, close, reopen and export. Listen to the exported sound and inspect the picture. Those are proposed acceptance checks, not completed lab work. They answer the creator's question more directly than the size of the Rust dependency graph.
The current README also documents a built-in MCP server through concat --mcp and a JSON-RPC interface. Developers evaluating automated editing should check the exact operations they need in a complete import-to-export workflow. A feature in today's documentation does not establish that it existed in the measured October 17 checkout.
Beta status and unsigned Windows packages narrow the audience
Concat's README labels the application beta. The v0.2.8 release, published October 20, 2026, says the Windows release is not code-signed and identifies Smart App Control as a restriction. If your workplace requires signed software, resolve that requirement before making this your editing application.
The repository API records a last push of October 28, 2026. The recent feed includes an October 30 macOS opening report and a proposal for live MCP edits. These show ongoing development and incoming reports. They do not establish response times or reliability. The release date alone would be a poor way to judge maintenance, especially when the measured checkout predates that release.
Shotcut and Kdenlive offer other desktop evaluation routes
Shotcut documents regularly available binary downloads. Its README reserves source building for beta testers or contributors who know that process. Consider its packaged editor if your immediate job is cutting footage and you have no reason to work on Concat's source.
Kdenlive documents stable releases and experimental daily builds. Its README identifies an MLT, Qt and KDE application stack, with primary development on KDE Invent. It is another desktop editor to evaluate if that distribution and contribution route suits your team. Neither alternative was measured alongside WolfCut, so there is no comparative speed or reliability winner here.
Our setup rating remains unassessed because the sandbox never completed the required native-library setup. Support responsiveness and operational maturity also remain unassessed. For a creator considering Concat, the next useful evidence is a saved project that reopens and exports correctly on the intended machine.