The idea started from a simple observation: thumbnails directly influence click-through rates, yet most creators spend disproportionate time designing them manually or rely on generic tools.
Thumbgen uses prompt-based image generation models (via OpenAI and Replicate) combined with template intelligence and real-time feedback to craft thumbnails optimized for visibility and engagement. The system is designed for creators who value speed, precision, and visual consistency.
Built with Next.js and Tailwind CSS, Thumbgen prioritizes performance and simplicity. There is no required design knowledge. Users simply describe what they want, and the system handles the rest.
Currently at 2,000+ concurrent users and growing, the platform is still evolving. We are exploring native YouTube integrations, A/B testing capabilities, and localized thumbnail adaptation.
Would appreciate feedback from the HN community regarding performance, infrastructure decisions, or use cases we might be overlooking.