mrkeyoor.comAsk

mayocream/koharutested 29 Sept 2026commit 988a6cd

koharu review

At commit 988a6cd, our source build failed when pkg-config could not find GTK4; Vitest passed 102 cases.

  • Install 460 s
  • Build failed
102/102tests passed
  • 1,544 packages
AI Toolsevaluationupdated 31 Oct 2026
+96stars / 7d

Koharu is a desktop manga translation editor with text detection, OCR, translation, cleanup and page composition. Its English README documents proofreading, a WebGPU canvas and layered PSD export. The requested mayocream/koharu repository resolves to koharu-rs/koharu.

Verdict

At commit 988a6cd, our source build failed when pkg-config could not find GTK4; Vitest passed 102 cases. Try the official release if you want to evaluate Koharu's documented correction and lettering workflow. We did not launch it or measure translation quality, so inspect a representative page and its exported layers before adopting it.

We ran it

Lab card: what happened when we ran koharu
Lab card from our sandbox run
Screenshot of koharu (koharu.rs)
koharu.rs
Install✓ · 460s1544 packages · 2143 MB
Build✗ · 336s
Tests✓ · 30s102 passed · 0 failed of 102 (vitest)
Repo783 files~212,717 lines of source · 47.3 MB · 6 CI workflows

Answers from our run

Does koharu build from source?

Dependencies installed in 460 seconds (1544 packages), and the build failed. We cloned commit 988a6cd into a clean Debian container with 3 CPUs and no project-specific setup.

Do koharu's tests pass?

Yes. The project's own test command finished in 30 seconds and reported 102 passed, 0 failed (vitest).

Who should not use koharu?

Source builders expecting Bun dependencies alone to supply the desktop environment: pkg-config could not find GTK4 in our build.

What are the alternatives to koharu?

Manga Image Translator, BallonsTranslator. At commit 988a6cd, our source build failed when pkg-config could not find GTK4; Vitest passed 102 cases.

Setup2/5Our source checkout installed but the build failed.Editor’s note: The source build failed at the GTK4 lookup; release installation and desktop launch were not reproduced.
Docs4/5On the visible editorial rubric, the supplied README and installation guide explain the main evaluation requirements; coverage across all documented paths was not independently checked.
Community4/5Last push on 2026-10-30 and 6,669 stars.Editor’s note: The supplied evidence does not establish user-support responsiveness or maintainer capacity.
Maturity3/5Latest release 0.83.5 on 2026-09-22 (pre-1.0), repository created 2025-04-09 and its test suite passed in our sandbox (102 of 102 passed).Editor’s note: Vitest passed the recorded suite, but desktop operation, model output and export behavior were not exercised.

How we score

Who it’s for

Manga translators who want to correct OCR and translated text alongside page composition.
Editors evaluating a layered PSD handoff for later image editing.
Developers prepared to maintain Rust, Bun, LLVM and native platform dependencies.

Who it’s NOT for

Source builders expecting Bun dependencies alone to supply the desktop environment: pkg-config could not find GTK4 in our build.
Users requiring an offline first launch without provisioning runtime packages and selected models.
Machines without a working WebGPU graphics environment, even when inference uses the CPU.
Teams requiring local processing while selecting a remote translation API.

Setup reality

Our September 29, 2026 sandbox run at commit 988a6cd installed 1,544 dependencies in 460 seconds and occupied 2,143 MB. The build failed after 336 seconds when pkg-config could not find GTK4. Vitest completed successfully in 30 seconds with 102 passes and no failures.

The log does not distinguish an absent GTK4 installation from a pkg-config search-path problem. The failed build prevented a finished desktop-app evaluation, including canvas initialization, model inference and export behavior.

The installation guide recommends a release build for translation work. First launch can download native runtime packages, and selected models download on first use. Source development has separate Rust, Bun, LLVM and platform requirements.

The recorded sandbox configuration had 3 CPUs, 8 GB RAM and image lab-node:22. That image label does not establish the actual runtime or operating-system version.

Try the release before taking on a source build

Koharu is worth evaluating if your manga workflow needs OCR correction, translation and page composition in one editor. Our source build at commit 988a6cd stopped when pkg-config could not find GTK4. We did not launch the desktop application or measure translation quality.

The installation guide recommends a release build for translation work and development setup for changing Koharu's code. For an editor assessing the workflow, the release is the sensible starting point. It still needs a trial on the intended machine and a representative page.

The English README documents selectable detection, OCR, translation and inpainting stages at page or project scope. It also describes proofreading, manual cleanup, text placement and layered PSD export. These are documented capabilities; their output quality was not established by our run.

What happened when we ran it

On September 29, 2026, our sandbox installed 1,544 dependencies in 460 seconds and used 2,143 MB on disk at commit 988a6cd. The build failed with exit code 1 after 336 seconds. Vitest completed successfully in 30 seconds with 102 passes and no failures.

The build log reports that pkg-config could not find gtk4. It says the required gtk4.pc file needs to be installed and its parent directory included in PKG_CONFIG_PATH, which was unset. This establishes a failed GTK4 lookup. It does not distinguish an absent installation from a search-path problem, or show a defect in OCR, translation or inpainting.

The recorded sandbox configuration had 3 CPUs, 8 GB RAM and image lab-node:22. The label does not establish an actual Node or operating-system version. The checkout contained 783 files and roughly 212,717 source lines, with monorepo workspaces. The dependency footprint describes that source workspace, not a downloaded release or its model collection.

The passing Vitest suite does not establish desktop launch, canvas initialization, inference quality or export behavior. Those remain untested in this run.

Downloads and graphics remain part of release evaluation

The installation guide describes release packages for Windows, macOS and Linux. It also tells Linux users to install desktop-library dependencies reported by their package manager, so choosing a release does not guarantee an environment without additional setup.

First launch can take longer while native packages download. Model files download when you first use the selected model. The guide names GitHub release assets, PyPI for CUDA and AMD's repository for ROCm as runtime sources, with model weights from Hugging Face. Prepare the selected assets before relying on an offline workflow.

The editor canvas uses WebGPU and requires a current graphics driver even when inference runs on the CPU. CPU inference therefore does not remove the graphics requirement. The README also documents Metal on Apple silicon and CUDA, ROCm/HIP and Vulkan on Windows and Linux, with compatibility depending on the selected path. Our run provides no inference results with which to rank them.

Translation can use a local language model or a remote API. Choose the processing route according to the project's privacy requirements before importing unpublished pages. Local inference and hosted translation are separate options.

Source contributors should follow the platform setup instructions linked from the README. Rust, Bun and LLVM are prerequisites; installing the Bun workspace alone did not establish that our native application could build.

Compare batch work with interactive correction

Manga Image Translator documents a local batch mode. Evaluate it when processing a folder through a batch command is your starting point.

BallonsTranslator documents mask editing and WYSIWYG rich-text editing. Compare it when repairing cleanup masks and adjusting lettering are central to your workflow. Neither alternative was measured against Koharu in our sandbox.

What the documentation rating means

Documentation receives 4 on this editorial scale: 1 means essential instructions are absent, 3 means a usable path has material gaps, 4 means the supplied instructions explain the main evaluation requirements but have not been checked across all documented paths, and 5 requires independently checked coverage across those paths. The English README and installation guide explain release selection, first-launch downloads, graphics requirements and the distinction between translation work and source development. That supports the documentation judgment, not a compatibility guarantee.

Setup ease, community support and finished-product maturity remain unassessed. We did not reproduce release installation and launch, inspect support responsiveness, or exercise model output and exports.

For your trial, choose a page with narrow speech bubbles and lettering over artwork. Run the selected stages, correct the text and cleanup, then inspect both the preview and exported layers. The adoption decision is how much correction that page still needs on your machine.

Alternatives

ProjectWhat it isPick it when
Manga Image TranslatorAn image translation pipeline with a documented local batch mode.Evaluate it when processing a folder through a batch command is your starting point.
BallonsTranslatorA comic translation editor with mask editing and WYSIWYG rich-text editing.Compare it when repairing cleanup masks and adjusting lettering are central to your workflow.

What people are saying

  1. [github-trending] koharu-rs/koharu

Sources

  1. Koharu English README
  2. Koharu installation guide
  3. Manga Image Translator README
  4. BallonsTranslator English README

More ai tools reviews