Back to Blog
Showcase3 min read

Watch: A Quick Intro to Lastest

Watch: A Quick Intro to Lastest

We made a short video walkthrough of Lastest so you can see the product end-to-end before you install anything. It covers the pieces that are hardest to convey in screenshots: how the AI generates tests, how the dashboard reacts to a failing build, and how the diff slider actually feels under your cursor.

What you'll see

The video walks through the core loop of visual regression testing with Lastest - from pointing the agent at a staging URL to reviewing a diff on a pull request. No slideware, no fake screenshots. Everything in the clip is the real self-hosted dashboard.

Flow diagram of the loop shown in the video: point the agent at a staging URL, AI writes Playwright tests, a human reviews and approves, the baseline locks in, and every PR replay runs free; AI only runs at create or fix time.
The video's core loop: AI runs once to author tests, then every approved baseline replays for free.
  • AI-generated test suites - point the agent at a URL, it crawls, clicks, and writes tests for you. You review and approve before anything becomes a baseline.
  • Three diff engines side-by-side - pixel, structural (SSIM), and perceptual (Butteraugli) comparison, each tuned for a different tolerance budget.
    • Layered stack of the three diff engines run on a single baseline-vs-current screenshot: Pixel (Pixelmatch) pixel-perfect and noisy, Structural (SSIM) DOM and layout-aware, and Perceptual (Butteraugli) human-eye aligned, each with its own tolerance budget.
    One screenshot, three diff engines stacked by tolerance budget, exactly as the video shows them side by side.
  • The diff slider - drag across a baseline/current pair to see exactly what changed. No squinting at red boxes.
  • Self-hosted dashboard - one Docker command, your data stays on your infrastructure, unlimited screenshots.

Why a video?

Visual testing is, by nature, visual. A blog post can tell you the diff slider is fast; a 30-second clip of someone actually dragging it is worth more than any paragraph. The same goes for the AI agent - watching it navigate a real app and capture a test suite makes the value obvious in a way that a bullet list never will.

We also wanted to answer the most common question we get from teams evaluating Lastest: "what does it look like day-to-day?" The video is our shortest honest answer.

Try it after you watch

If the video makes you curious, the fastest way to go further is to spin up the dashboard yourself. It's one Docker command, no signup, no credit card:

docker run -p 3000:3000 ghcr.io/las-team/lastest:latest

From there you can point it at any staging URL and let the AI agent generate your first test suite. Your first run uses AI tokens to build the suite; every replay after that is free, forever.

Stat cards comparing what Lastest costs after the video: $0 self-hosted forever with unlimited screenshots, $0 Cloud Free evaluation tier, and $299/mo flat Cloud with no per-seat fees, plus a banner stating every replay after the first build costs 0 tokens forever.
After you watch: self-hosting is free forever, Cloud is a flat $299/mo, and every replay costs zero tokens.

If you'd rather read than watch, the docs cover the same ground in more depth. And if you have feedback on the video - or want to see a longer deep-dive on a specific feature - let us know on GitHub.