Local image optimization (no uploads, no account)

1. Quick summary

• 100% local: Canvas + built‑in encoders, zero upload.
• Quality slider: 30–95 to balance size & fidelity.
• Auto format: Chooses JPEG/WebP for best bytes.
• Batch flow: Optimize N images sequentially.
• Multi download / ZIP: >5 files packed automatically.
• Side‑by‑side compare: Opens a preview window.

2. Recommended workflow

1. Load images (JPG, PNG, WebP).
2. Set global quality (start 80 photos, 75 illustrations).
3. Keep auto unless forcing a specific format.
4. Click Optimize all & inspect savings.
5. If artifacts (banding, blocks) appear, raise quality by 5.

3. Internal mechanics

Each file is decoded to an <img>, drawn on a <canvas> and re‑encoded via canvas.toBlob() as image/webp or image/jpeg. No bytes leave the browser; ObjectURLs are revoked post‑download.

4. Format decision

Auto prefers WebP for PNG (esp. transparency) and evaluates JPEG photos; if WebP doesn’t win clearly, JPEG remains. Typical WebP savings: 20–35% vs JPEG.

5. Performance & limits

• O(n) in pixel count.
• Medium images (≤3000px longest side) in milliseconds.
• Very large (>6000px) may consume memory/time.
• On memory pressure, split into smaller batches.

6. Privacy

No transfer, ideal for confidential or NDA material. Business model: lightweight ads, not data harvesting.

7. Quick checklist

• Final weight <250KB hero / <100KB inline.
• Dimensions resized before compression.
• Modern format (WebP) when bytes saved.
• No perceptible artifacts.

8. Roadmap

Upcoming: AVIF, smart presets, local SSIM/PSNR metrics.