fieldcards: a local Gemma turns your photos into a printed scavenger hunt

Description and use

fieldcards turns outdoor photos into printable scavenger‑hunt cards using a local Gemma 3 model via Ollama, with no internet upload and quick processing.

Problem

Generate printable scavenger‑hunt clues from outdoor photos without uploading images or using an API.

Outcome

fieldcards turns a folder of outdoor photos into a printable scavenger‑hunt sheet in about two minutes for the first photo and 7‑13 seconds per subsequent photo, using only local resources and no internet.

Workflow

• 1. Install Gemma 3 locally with Ollama and the fieldcards npm package.
• 2. Run fieldcards ./photos --title "Saturday in the park" to process a folder of photos.
• 3. The tool posts each image to the local Ollama server with a JSON schema prompt.
• 4. Gemma returns a structured response (subject, clue, etc.).
• 5. The response is validated, clues are cleaned, and ASCII cards are generated using altsvg.
• 6. Output files are written: index.html, cards.txt, deck.json.
• 7. Users can print the sheet or re‑run with --from deck.json to edit without re‑running the model.

Evidence

Each photo goes to Gemma 3, running locally through Ollama. Gemma works out what the photo shows and writes a clue for finding it that doesn't name it.

Source

Every run writes three files: index.html, cards.txt, deck.json.

Source

On my laptop (RTX 3050, 4 GB), the first photo takes about two minutes because it also loads the model. Each photo after that takes 7 to 13 seconds, using about 2.9 GB of GPU memory.

Source