mrkeyoor.com_
Wed 23 Sept 06:49 UTC
AI Toolsevaluationupdated 23 Sept 2026

shrimply review

Shrimply is a pre-alpha desktop video editor for NVIDIA Linux systems and Apple Silicon Macs, with timeline editing, compositing, captions, GPU export, AI media tools, and an MCP control surface. It aims at work that normally spills across a video editor, motion-graphics package, caption tool, and small automation scripts.

Verdict

Our Shrimply checkout pulled 858 packages, then both build and test stopped because the system had Pango 1.50.12 instead of the required 1.57. Treat Shrimply as a compelling codebase to study or help build, not as the editor holding your only copy of a client project. Its feature range is real, but the project's own pre-alpha warning, narrow hardware support, and missing binary release make experimentation the responsible use today.

We ran it

Lab card: what happened when we ran shrimplyScreenshot of shrimply (shrimply.pages.dev)
Install✓ · 82s858 packages
Build✗ · 65s
Tests✗ · 24sran, no count parsed
Repo2065 files~397,561 lines of source · 51.7 MB · 4 CI workflows · Dockerfile

Answers from our run

Does shrimply build from source?

Dependencies installed in 82 seconds (858 packages), and the build failed. We cloned commit a118985 into a clean Debian container with 3 CPUs and no project-specific setup.

Do shrimply's tests pass?

The test command failed in our container, and its output did not report a pass or fail count.

Who should not use shrimply?

Editors protecting paid work: the README warns about irreversible project-file breakage, crashes, resource leaks, and performance regressions.

What are the alternatives to shrimply?

Kdenlive, Shotcut, Blender. Our Shrimply checkout pulled 858 packages, then both build and test stopped because the system had Pango 1.

Setup1/582-second install; build needs newer native libraries and CUDA
Docs4/5Detailed editing, export, platform, build, MCP, and security guides
Community3/51,137 stars with platform issues active through September 22
Maturity1/5Pre-alpha, no tagged release, and project breakage is acknowledged

Who it’s for

Rust and graphics developers who want to contribute to an unusually broad video-editor codebase.
NVIDIA Linux or Apple Silicon users willing to test pre-alpha editing workflows on expendable project copies.
Motion designers interested in one timeline that can include ordinary media, Manim scenes, captions, paint, and 3D elements.
Automation developers exploring MCP-driven edits, transcription, text-to-speech, and generated video inside an editor.

Who it’s NOT for

Editors protecting paid work: the README warns about irreversible project-file breakage, crashes, resource leaks, and performance regressions.
AMD, Intel GPU, or Windows users expecting current support: open issues 54, 55, and 56 are still port requests, while issue 1 says cross-platform support is largely absent.
Debian builders with Pango 1.50: our build and test commands both stopped because the Rust GTK binding required Pango 1.57 or newer.
Anyone wanting a normal binary release: issue 36 says downloadable binaries should wait until the project is stable, and GitHub has no tagged release.
Export workflows that require CPU encoding: the documentation says H.264 and H.265 use NVENC with no software fallback.
Users opening untrusted Blender or Manim files: Shrimply warns that these can run arbitrary code with the user's account permissions.

Setup reality

Our sandbox installed 858 packages in 82 seconds. The build failed with exit code 101 after 65 seconds, and tests failed with exit code 101 after 24 seconds. Both logs stopped at the same check: installed Pango was 1.50.12, below the required 1.57.

Source development targets Fedora and needs Rust, native GTK and media libraries, submodules, Slang, CUDA 12.9, and about 70 GB of free disk according to the project docs. A Nix shell and an experimental Docker build path are also documented.

The measured checkout had 2,065 files, about 397,561 lines of source, 4 CI workflows, and a Dockerfile. The README limits normal use to an NVIDIA or Apple Silicon GPU on Wayland or macOS 15 or newer, and labels the application pre-alpha.

Shrimply is an editor you can automate, not a safe daily driver

Shrimply combines a multitrack video editor with a large set of media and automation experiments. A new .shrimp project starts with caption, video, and audio tracks. The editor has preview controls, an inspector, timeline trimming, ripple cuts, snapping, grouping, and undo history. It can open its own format plus JSON, OpenTimelineIO, and Kdenlive projects. The repository's warning matters more than that list: the application is pre-alpha and may corrupt project files.

The unusual part is how much sits behind the timeline. The source tree includes transcription, text-to-speech, lip sync, stabilization, paint, layered images, Manim, Blender, video generation, and an MCP integration. Its MCP tools can inspect editor state and apply typed edit operations as one undoable action. That breadth makes Shrimply interesting to contributors. It also explains why the checkout has about 397,561 lines of source before dependencies.

MCP access is tied to the open editor rather than a second project model. Resources expose editor state, project clips, individual clip details, and the typed edit API. A script can validate an ordered batch against a cloned project, then install the batch as 1 undoable history action. The same interface can insert text, configure speech generation, transcribe addressed audio clips, or import files with explicit collision behavior. That is a credible automation surface even while the desktop application remains risky.

NVIDIA Linux and Apple Silicon are the practical targets

The README requires an NVIDIA or Apple Silicon GPU and lists Wayland or macOS 15 and newer. Linux development targets Fedora, while the checked-in Nix environment targets x86_64-linux and still needs a compatible NVIDIA driver. The build documentation specifies CUDA Toolkit 12.9 and says the current package embeds kernels for compute capability sm_86. Those are narrower requirements than the phrase cross-platform video editor suggests.

Open issue 1 states that the project has little cross-platform support and proposes moving more rendering work toward Slang or WGPU. AMD and Intel support each have an open port request, and WinUI 3 has another. A macOS 15 user also reported a v0-era build crashing because an interface class available on newer macOS was missing. These reports do not prove that every current machine fails, but they make platform testing a first step rather than a final check.

What happened when we ran it

Our sandbox installed commit a118985 in 82 seconds and pulled 858 packages. The repository itself occupied 51.7 MB, with 2,065 files and roughly 397,561 source lines. The build then failed after 65 seconds with exit code 101. This happened in an unprivileged Rust container with 3 CPUs and 12 GB of RAM.

The error came from a native library check. pkg-config found Pango 1.50.12, while pango-sys required Pango 1.57 or newer. The log also said PKG_CONFIG_PATH was unset, but it had already identified the installed Pango version and declared it too old. We cannot turn that into a Rust compiler finding because compilation stopped at the system dependency gate.

Tests failed after 24 seconds with the same exit code 101 and the same Pango requirement. The supplied log showed no passed or failed test count, so none should be inferred. Our scan found 4 CI workflow files and a Dockerfile, but no top-level tests directory. The failed local run still matters: the project documentation's Fedora target includes native versions that a fresh Debian-based environment may not provide.

The editor already covers serious timeline work

Shrimply can export H.264, H.265, GIF, several caption formats, and JSON project data. The 2 video codecs require NVENC and have no software encoder fallback. Caption export includes SRT, WebVTT, ASS, YTT, and plain text, with format-specific handling for styling and timing. The Kdenlive importer preserves some track and media details but explicitly omits transitions, track effects, title clips, and other unsupported MLT features.

That honesty is useful. It tells an editor which round trips are impossible before a project is moved. Security documentation does the same for executable sources: Blender and Manim files can run code with the user's permissions. Shrimply asks the user to trust individual files or folders, then remembers that choice outside the project. The approval is not a sandbox or malware scan, so shared projects deserve a careful source review.

September activity is strong, while releases are still absent

GitHub recorded the last push on September 19, 2026. The repository had 1,137 stars and 15 open issues and pull requests when fetched, and platform discussions were updated through September 22. GitHub returned no latest release. Issue 36 explicitly ties downloadable binaries to reaching a safer level of stability, which matches the README's pre-alpha label.

Shrimply is easier to recommend as a contribution target than as a production tool. The 858-package install and Pango 1.57 floor show the cost before CUDA, codecs, and GPU behavior enter the picture. If its MCP editor control or mixed media timeline solves a problem you cannot test elsewhere, use disposable projects and keep exports outside the .shrimp file. For routine work, Kdenlive or Shotcut asks you to accept far less risk.

Alternatives

ProjectWhat it isPick it when
KdenliveA mature open-source nonlinear editor built on MLT and KDE frameworks.pick this instead when dependable everyday editing and broad desktop support matter more than Shrimply's experimental AI and MCP tools.
ShotcutA cross-platform GPL video editor with conventional timeline and export workflows.pick this instead when you need an installable editor across Windows, macOS, and Linux without a CUDA-focused source build.
Blender gh↗A 3D creation suite that also includes compositing and a video sequence editor.pick this instead when 3D and compositing depth outweigh the appeal of Shrimply's newer editor and automation model.

What people are saying

  1. [velocity-scout] soirihiroka/shrimply

Sources

  1. Shrimply repository
  2. Shrimply getting started guide
  3. Shrimply development guide
  4. Shrimply export guide
  5. Cross-platform support issue
  6. Binary release issue
  7. macOS 15 launch report
  8. Shrimply security guide

More ai tools reviews

reverify · course2md · interdimensional-game · infinite-livestream · bkn-foundry · tokenizers · the whole board →