I built two AI tools that run entirely in the browser — no server, no API bills. Here’s what worked and what didn’t.

Description and use

The author added OCR and image upscaling to a browser‑only tool, eliminating server costs. Background removal failed due to loading issues.

Problem

Add AI-powered features to a free converter toolkit without incurring API costs or maintaining a backend.

Outcome

Successfully added OCR and image upscaling to a browser‑only tool, eliminating server costs. Background removal was omitted due to reliability issues.

Workflow

• 1. Integrated Tesseract.js for OCR, running fully in-browser.
• 2. Added UpscalerJS for image upscaling, leveraging WebGL for acceleration.
• 3. Attempted background removal with a CDN‑hosted model, but faced CORS, timeout, and inconsistent loading issues.
• 4. Decided to drop the background removal feature after repeated failures.
• 5. Documented the build process and shared code for replication.

Evidence

OCR via Tesseract.js — genuinely solid text extraction, runs fully in-browser

Source

Image upscaling via UpscalerJS — WebGL-accelerated, no server round-trip

Source

I also tried adding background removal using a model pulled from a CDN. Looked simple on paper. In practice, the model failed to load reliably across browsers — CORS issues, CDN timeouts, inconsistent behavior.

Source