Price terms as published
Each line is the vendor’s own wording, split from the price capture checked on 2026-09-21. Nothing here is rounded, converted or re-worded — the full sentences stay in the review below. Lines that carry only a price continue the plan named on the line above, in the vendor’s own order.
| # | Plan / term | Checked |
|---|---|---|
| 1 | Free (MIT) | 2026-09-21 |
| 2 | hosted demo with API key | 2026-09-21 |
One-liner: screenshot-to-code is an open-source app that takes a screenshot, design mockup, or short screen recording and generates a working front-end implementation — plain HTML with Tailwind CSS, or React/Vue components — using multimodal models.
Verified facts (checked 2026-09-21)
| Attribute | Value | Source |
|---|---|---|
| GitHub stars | 79,442 (9,706 forks) | abi/screenshot-to-code via GitHub API, 2026-09-21 |
| License | MIT | GitHub API |
| Owner | abi (Abi Raja) | repository metadata |
| Last push | 2026-09-09 (1,455 commits; hosted demo at screenshottocode.com) | GitHub API / GitHub page |
| Output stacks | HTML + Tailwind, React, Vue; screen-recording input supported | repository README |
| Model configuration | Self-host with your own OpenAI / Anthropic / Gemini / Replicate keys (.env based) |
repository README |
What it actually does
You drop in an image of a UI — a sketch, a Dribbble-style mockup, or a photo of an existing screen — and the system sends it to a vision-language model with a coding prompt, then renders the returned code in a live preview so you can compare it side by side with the source image. Video mode reads a short recording of interactions and attempts to reconstruct the behavior, not just the static layout.
The repository includes a hosted demo, but running it yourself is straightforward: the back end accepts standard OpenAI-compatible API configuration, so you can point it at the model provider of your choice rather than relying on the public instance.
What the marketing doesn’t say
The generated output is a first draft. Simple layouts and common component patterns come back impressively close to the source; complex design systems, exact spacing, accessibility semantics, and real data wiring usually need a developer to finish. The quality ceiling moves with the model you plug in — the repository is the harness and UI, not a model itself, so claims about output accuracy are really claims about whichever vision model you configure.
Because it is MIT licensed, there are no copyleft or commercial-use restrictions on the code you generate or on embedding the tool in your own workflow.
Who should use it
Front-end developers who want to skip the mechanical first pass of turning a picture into markup, and solo founders prototyping landing pages from sketches. If you need pixel-perfect production code with no review step, no screenshot-to-code tool — this one included — honestly delivers that yet.
Repository stats and license verified against primary sources on 2026-09-21. This entry follows our verification policy: every claim here was checked against primary sources, not copied from other reviews.