Hugging Face puts vision models in a spreadsheet—no code needed

Hugging Face just dropped a massive update to AI Sheets, its open-source tool for enriching data with AI. The headline feature? Vision support. You can now drag in a folder of images—receipts, handwritten notes, product photos—and apply vision models as easily as you’d run a formula. The tool runs on Inference Providers, giving you access to thousands of open models from the best providers out there without writing a single line of Python.

The workflow is built for people who live in spreadsheets. You upload your data, click a column header, and define an AI action with a prompt and a model. For image columns, that means extracting text from scanned documents, classifying images, or pulling structured data like merchant names and totals from a stack of receipts. The interface has a feedback loop baked in: a thumbs-up on a good result teaches the model what you want, using your approvals as few-shot examples to improve future generations.

Hugging Face isn’t just stopping at extraction. You can generate and edit images in the same sheet. The example they give is a content calendar for social media. You’ve got post titles and descriptions. Create an image column with a prompt like “Generate an appetizing food photo for: {{title}}” and the sheet populates with visuals. Don’t like the background? Add another column to edit it. The whole thing—copy and visuals—stays in one place, ready for export as CSV or Parquet.

This feels like the logical endpoint of the “no-code AI” trend. It’s not a flashy demo; it’s a utility. The open-source angle is the real differentiator. You’re not locked into one vendor’s model or pricing. If a better model drops tomorrow, you swap it in. For anyone who’s ever manually transcribed a hundred receipts or begged a designer for 50 thumbnail variations, this is genuinely useful infrastructure masquerading as a spreadsheet add-on.