Roborazzi vs Visual Regression Tracker

Roborazzi is an open-source library for screenshot testing in JVM-based Android integration tests. It records images, compares them with saved baselines and can fail verification when a screenshot differs. Through Robolectric, Android tests can run as JVM tests without a device or emulator. Capture options include Jetpack Compose nodes, Espresso views, Android Views and Compose lambdas; the library can also create screenshot tests for Compose previews and common Compose Multiplatform previews. It supports Compose Desktop and JVM preview tests, rendering images with the host’s Skia, while Compose Multiplatform iOS support is experimental. Gradle tasks cover recording, comparison and verification, and the project documents GitHub Actions workflows for screenshot checks in CI. An IDE plugin shows test screenshots in IntelliJ IDEA and Android Studio. Optional capabilities include UI-tree JSON output, accessibility checks, animated GIF or APNG capture, and experimental AI image assertions. Roborazzi is free, available on Maven Central and licensed under Apache License 2.0. It has no direct Bazel support, and graphics may render differently across operating systems.

Visual Regression Tracker compares submitted screenshots with accepted baselines to flag unexpected visual changes. It keeps baseline history and lets teams exclude image regions from comparisons. Projects can choose among Pixelmatch, Looks-Same, Odiff, and a VLM comparison provider. The review dashboard supports side-by-side, diff, scrubber, and onion-skin views, with one-click approval. Feature branches inherit the main branch baseline until a change is approved, and builds can post commit status to GitHub or GitLab. The project lists agents for Playwright, Cypress, CodeceptJS, and Robot Framework, plus JavaScript, Java, Python, and .NET SDKs; its REST API supports other languages. A standalone JAR can compare PDFs as well as images. Deployment is self-hosted using Docker images, and the project says this keeps data inside the user's intranet. The documented screenshot upload payload limit is 20 MB. Each project has one owner; members can view snapshots and approve changes, while some actions are owner-only. It is open source under Apache-2.0.