Clean. Detect. Generate. In Seconds.
RemoveBanana is an all-in-one toolkit that removes the hard parts, so you can clean, detect, and generate in seconds.
- Remove Gemini and Imagen watermarks from images
- Extract clean MP4 links from shared Sora videos
- Remove image backgrounds locally in the browser
- Detect AI-generated images with a hybrid voting pipeline
- Generate styled QR codes for multiple payload types
Live site: https://removebanana.aashuu.tech
| Tool | Route | What it does | Processing model |
|---|---|---|---|
| Gemini/Image Watermark Remover | /image-remover | Removes Gemini/Imagen/Nano Banana watermarking using reverse alpha blending | Client-side |
| Sora Video Remover | /video-remover | Accepts Sora links and extracts clean MP4 + prompt | Server route fetch + parsing |
| Background Remover | /background-remover | Removes background from one or many images, supports recolor/recompose and ZIP export | Client-side AI model |
| AI Image Detector | /ai-image-detector | Multi-engine AI-image detection with weighted voting and JSON export | Server route + external APIs + metadata analysis |
| QR Generator | /qr-code-generator | Styled QR generation for URL/text/Wi-Fi/vCard/event/phone/SMS/geo with PNG/JPEG/SVG export | Client-side |
- Uses reverse alpha blending (see src/lib/watermark.ts)
- Auto-selects watermark config (48px vs 96px pattern)
- In-browser processing for private local workflow
- Supports public Sora links
- Handles both /p/ shared links and /g/gen_* pattern with fallback extraction
- Returns extracted prompt and direct MP4 URL for download
- Multi-file queue with per-card status
- Local model warm-up and progress tracking
- Output controls: transparent background, solid color, custom image background
- Single image PNG download + ZIP batch export
- Ensemble of: Sightengine GenAI classifier, Hugging Face inference model, and local metadata heuristic analysis
- Weighted scoring with thresholding
- Full downloadable JSON report from UI
- Payload types: text/URL, Wi-Fi, vCard, event, phone, SMS, geo
- Style controls: gradients, dot styles, corner styles, logo upload, randomizer
- Export: PNG, JPEG, SVG
- Extra actions: copy image, embed code, print
- Most heavy media workflows are browser-side by design
- Install prompt uses localStorage-based dismiss window
- Service worker is pass-through (installability-oriented)
- Background remover route has COOP/COEP headers via next.config.ts
- Centralized brand metadata in src/lib/brand.ts
- Rich metadata in src/app/layout.tsx
- robots.ts + sitemap.ts configured for production domain
- Web app manifest via src/app/manifest.ts
