HackerTrans
TopNewTrendsCommentsPastAskShowJobs

h2bomb

no profile record

Submissions

Show HN: A privacy-first bulk image compressor and HEIC converter in the browser

zip.easynote.cc
2 points·by h2bomb·6 เดือนที่ผ่านมา·2 comments

Show HN: Gemini Watermark Remover – A web tool using reverse alpha blending

re.easynote.cc
2 points·by h2bomb·7 เดือนที่ผ่านมา·1 comments

comments

h2bomb
·6 เดือนที่ผ่านมา·discuss
I built a simple, fast, and privacy-focused web tool for bulk image compression and HEIC conversion.

The Problem: Many online compressors upload your photos to their servers, which is a privacy risk and slow for large batches.

The Solution: This tool handles everything 100% locally in your browser.

Technical Stack:

image-compression: Uses Web Workers for non-blocking UI during heavy compression tasks.

heic2any: Enables browser-side HEIC/HEIF to JPEG conversion (handy for iPhone users).

jszip: Batches processed images into a single ZIP for one-click download.

Concurrent Processing: I implemented a custom queue (MAX_CONCURRENT = 3) to balance speed and browser stability.

Key Features:

No server uploads (Privacy-first).

Batch processing with ZIP export.

Adjust quality and format (WebP/PNG/JPG) on the fly.

Dark mode support & Responsive UI.

I'd love to hear your feedback on the processing performance or any format support you'd like to see!

Site: https://zip.easynote.cc/
h2bomb
·7 เดือนที่ผ่านมา·discuss
I’ve been using Google's Gemini for image generation, but the watermarks are often a hurdle for clean designs.

I built this web tool based on the Reverse Alpha Blending principle. Unlike typical "AI erasers" that use generative inpainting to fill in gaps (which often results in blurriness), this tool treats the watermark as a standard alpha-composited layer and mathematically reverses the blending formula.

Key Technical Implementation:

Pure Client-Side: The core engine is written in vanilla JavaScript. It uses the Canvas API to extract ImageData and processes pixels directly in the browser. No images are ever uploaded to a server.

Deterministic Restoration: It reverses the formula: $original = (watermarked - \alpha \times 255) / (1 - \alpha)$. By using a pre-calculated Alpha Map from the Gemini logo, it achieves pixel-perfect restoration with zero AI "hallucinations".

Auto-scaling Logic: The tool automatically detects whether the image uses the 48px or 96px watermark variant based on the source dimensions (1024px threshold), ensuring the correct $\alpha$ mask is applied.

Optimization: It uses Float32Array for the alpha map and implements threshold clamping (e.g., ignoring $\alpha < 0.002$) to avoid floating-point noise and division-by-zero errors.

I'm standing on the shoulders of giants here—special thanks to Allen Kuo for the original algorithm research and CLI tool [0]. I wanted to make this technique accessible to non-technical users through a frictionless web interface.

Web Version: https://re.easynote.cc Original Research: https://github.com/allenk/GeminiWatermarkTool