Skip to main content
Back to blog
Product
·5 min read

Introducing Vision Credits: AI That Sees Your Photos and Videos

Talyx has always been great at classifying documents — PDFs, contracts, spreadsheets. Drop a folder, and the AI reads the content to sort and rename everything. But images? Videos? Those got generic names like photo_20260330.jpg. Not anymore.

The Problem: AI Can't Read Pixels Through Text

Our Claude Code SDK integration sends text to Claude for classification. It works brilliantly for documents — Claude reads the content, understands it's an invoice or a contract, and names it accordingly. But images don't have text content. A photo of a cat is just a pile of pixels. Without actually seeing the image, Claude can't tell a cat from a car.

The Solution: Vision Credits

Vision credits give Talyx the ability to see your images and videos. When you sort a folder containing photos, Talyx creates a small thumbnail (512px max, about 50KB) and sends it to our secure server. There, Claude's multimodal vision analyzes what's in the image and returns a descriptive name.

The results speak for themselves:

  • image-5.jpgwhite-persian-cat-sofa.jpg
  • image-6.jpgretro-orange-muscle-car-showroom.jpg
  • image-9.jpgtaj-mahal-front-view.jpg
  • video-1.mp4golden-retriever-playing-park.mp4

How It Works

The sort flow now has three AI stages:

  1. Documents — classified locally via your Claude subscription (free, your files never leave your machine)
  2. Images & Videos — small thumbnails sent to our secure server for Claude vision analysis (uses vision credits)
  3. Fallback — anything left gets a name based on file type and metadata

Pricing

Every Personal license ($29) includes 500 vision credits. Each image or video frame costs 1 credit. Documents are always free.

Need more? Buy 500 credits for $5, or subscribe to Talyx AI ($9/month) for unlimited vision on all file types — no Claude subscription needed.

Privacy

We take privacy seriously. Image thumbnails are processed in memory on our server and immediately discarded — we never store your images. The Anthropic API (which powers the vision analysis) does not use API data for training. Documents classified via the Claude Code SDK never touch our servers at all.

What's Next

Vision credits are available now. Update to the latest version of Talyx, check your credit balance in Settings, and try sorting a folder with photos. You'll see the difference immediately.