WordRemover：我用 Next.js 開發 AI 影片文字移除工具，徵求意見
Show DEV: I built an AI video text remover in Next.js (wordremover.com) — would love your feedback!
獨立開發者以 Next.js 建成 WordRemover，這是一款免費、在瀏覽器運作的 AI 工具，可移除影片中的硬編碼文字、字幕、時間戳及浮水印。工具使用逐影格背景重建模型，旨在保留原始影片解像度；開發者正徵求用戶對介面互動、重建效果及效能的意見。
Hey DEV community! 👋
I'm an indie developer and I recently built WordRemover — a free, browser-based AI tool designed to erase hardcoded text, subtitles, timestamps, and watermarks from videos directly in your browser.
As someone leveraging modern AI coding tools (Cursor & Claude Code) to handle full-stack development, I got the core workflow working with Next.js, Tailwind, and Supabase. However, I know there is HUGE room for improvement, and I'm here to humbly ask the senior engineers, UI designers, and video tech creators in this community to roast my project! 🛠️
🛠️ Tech Stack & Architecture
- Frontend: Next.js, Tailwind CSS, Canvas API (for interactive brushing)
- Backend & Auth: Supabase
- Hosting: Vercel
- AI Core: Frame-by-frame background reconstruction model optimized to preserve original video resolution.
🎯 4 Specific Things I Need Your Advice On:
I would be forever grateful if you could test wordremover.com and give me your honest, unfiltered feedback:
- UI/UX & Interactions: Does the brush overlay and timeline navigation feel responsive? How can I make the mask selection feel more intuitive?
- AI Reconstruction Edge Cases: In what specific video scenarios (e.g., complex motion, dense gradients, low contrast) did the AI background recovery fail or look artifact-heavy?
- Performance & Latency: Any architecture tips on optimizing web canvas rendering or reducing video encoding/decoding overhead?
- Codebase & Product Advice: If you were rebuilding this from scratch, what's one thing you would architect differently?
🔗 Try It Out Here
Don't hold back — brutal honesty, UI roasts, and technical suggestions are all warmly welcomed in the comments below!
Thank you so much for your time and wisdom! 🙏
來源：DEV Community · dev.to