Cerul

Cerul is video memory infrastructure that helps AI systems and people find the exact moment something was said or shown in a video. It indexes speech, on screen text and key visual frames, then answers plain language questions with timestamped results, each carrying evidence of what was found, where and at which second. It comes as two products that share one memory: Cerul Desktop and a Cloud API. Cerul Desktop is a macOS app for Apple silicon. You point it at your own folders, channels and podcasts, press Alt and Space, ask a question and jump to the moment. Local search needs no account, and media, transcripts and search data stay on your machine unless you explicitly allow cloud execution. The Cloud API is for developers building agents and products: a single POST to a search endpoint returns citable hits, and the app, MCP and API draw on one shared credits balance. Windows and Linux versions are on the roadmap. The desktop download is free. The Cloud API starts with 500 free credits once, with one concurrent job and no expiry. Top ups are prepaid with a $10 minimum at $0.02 per credit, allow five concurrent jobs, and expire after 12 months. Video indexing costs 2 credits per minute, OCR 0.1 credits per processed frame, search 0.5 credits per request ($10 per 1,000 searches), and semantic annotation 15 credits per minute with separate activation. Storage is free for 30 days per asset, then 2 credits per GB per month, billed daily. The release notes refer to a Go tier for the app, but its price is not shown on the pricing page. Unpaid storage pauses service and can lead to deletion after notice.