PhotoCraft Review: A Hands-On Look at the Web Editor

Review date: 18 October 2026. Tested version: the unmodified v0.12.0 WebAssembly package hosted on this site. This is our own host's workflow review, not an independent security audit or a review of every newer desktop release.

The short verdict

PhotoCraft is worth trying for a small local browser edit and for exploring an open-source, Photoshop-style workflow. Our reproducible test covers opening a PNG, changing pixels with a brush, exporting a PNG and saving and reopening a PSD. It does not establish professional Photoshop parity, complex PSD fidelity, desktop performance or reliability on every device. The Photoshop comparison puts those limits alongside the workflow.

How we tested import, editing and PNG export

We used Chromium and a 320 × 320 PNG test image. The test drops the file onto the actual upstream canvas, paints a stroke, exports a PNG through the editor's File menu, decodes the image and checks that pixels changed while its dimensions remained 320 × 320. This checks an actual edited download rather than only whether an interface appeared.

You can repeat the workflow in the editor on our Photoshop clone guide or on the homepage. Start with a non-sensitive copy, wait for the canvas to load, open the image and make an obvious mark before exporting. The hosted HTML, JavaScript and WASM are traceable through the source and provenance guide.

PSD save, reopen and undo

The same test saves the edited image as PSD, checks the file signature and dimensions, then drops that PSD back onto the editor. A second PNG export is compared with the first edited export at pixel level. In this small round trip, the reopened output matches. A separate workflow test checks that undo restores the original pixels after visiting help in another tab.

This is one simple document, not a corpus of layered professional designs. It does not cover every smart object, font, adjustment or blend mode. Use the PSD support guide to plan checks for your own files, including reopening the output in the receiving app.

Privacy, saving and device limitations

During this workflow, application asset requests are local-site reads, and no document-upload API is used. Known Google Analytics 4 and self-hosted Plausible usage requests are measured separately: local image processing does not make browsing anonymous or eliminate analytics. The safety guide explains that distinction.

The pinned web release has no autosave or crash recovery. Save before refreshing or switching renderers. Large files depend on device and browser memory; a desktop-sized screen is easier for the original English-only editor UI. Phone viewport checks do not prove support across Android browsers or graphics drivers.

Who should try it, and who should test more carefully?

Try it if you want to explore local browser editing without an account. For deadline-sensitive production, detailed typography, plugin workflows or complex Photoshop handoff, keep a proven tool available and judge using your own output checks. This hosted build has no generative fill or prompt editing. For native packages and current upstream releases, use the download guide; their features and version numbers are separate from this review.

Repeat the test instead of relying on the verdict

The downloadable files in the PSD guide let you inspect the input, saved document and rendered outputs. Start with the original synthetic PNG and make your own visible brush mark. Your hand-drawn stroke will differ from the automated stroke, so compare the reopened result against your own edited export. Do not use our edited image as the only reference for a different edit.

Our test uses the real upstream canvas rather than substituting a mock editor. It checks the output file format and dimensions, verifies that editing changed pixels and compares the PNG exported after reopening the PSD. Those checks are stronger evidence for this small workflow than a screenshot of a loaded interface. They still apply only to the stated input and operations.

Interpret a zero pixel difference narrowly

The published round-trip record contains the measured difference between the edited PNG and the PNG exported after saving and reopening the simple PSD. A zero result means those two outputs match for that run. It does not mean every layer type, effect or font is preserved, nor that Photoshop would render or edit the file identically.

We did not test a professional multilayer corpus, a color-managed print workflow or a maximum document size. The browser test environment can use software graphics; its success is not a speed comparison against a native application. A useful review carries these boundaries with the positive result rather than hiding them below a general compatibility score.

Evaluate the session controls as well as the canvas

The website adds help links, renderer choices and fullscreen controls around the unmodified editor. Existing workflow checks cover keeping an active session while help opens separately, and undo behavior after returning. A new editor tab is a separate session, however, and renderer changes restart the application. Save explicitly before leaving or changing the environment.

The website loads GA4 and self-hosted Plausible directly for visit measurement, without an analytics permission panel. Measurement is separate from the editor: the integration does not send document contents or local filenames. These are website behaviors rather than new engine features. The privacy policy identifies the services involved and explains that local image processing is different from ordinary hosting and measurement requests.

Check your own acceptance criteria

Write down the output you need, the features that must remain editable and the application that will receive the file. Run a small representative copy through that path and inspect the result outside the editor. A successful PNG export can satisfy a simple image task while being insufficient for a layered production handoff.

If an important requirement fails, keep the original and a known-good alternative workflow. The appropriate conclusion is that this version did not pass that requirement, not that every part of the project is unusable. Equally, a small passing test should not become a promise of universal reliability. Our feature and comparison guides help turn the review into a task-specific decision.

Report a result with enough context to reproduce

Include the pinned release, browser and operating system, renderer, input and steps. Identify whether the difference appears in the canvas, exported PNG or editable document. Use synthetic material when reporting publicly, and inspect filenames, logs and screenshots for private information. The community guide explains the upstream-versus-website support boundary.

This review is written by the maintainer of the host being reviewed. It is transparent first-party workflow documentation, not an independent endorsement. Dates identify the published revision and test record; they do not certify every newer desktop release. Use the samples and method to reach your own conclusion, and send a correction if a documented observation cannot be reproduced under the stated conditions.