Ollama's Free AI Can Now See Photos and Listen to Audio

Ollama, a popular free program for running AI models locally, released a new version that lets its version of Google's open-source Gemma 3 model understand images and audio, not just text. This is a big deal for privacy: every picture and sound file you share stays on your own computer. Nothing gets sent to a company's servers. You can chat with the AI while showing it a photo of a broken appliance, or upload a recording of a meeting and have it produce notes.

Under the hood, the update is surprisingly flexible. For images, the AI uses special vision systems to make sense of what's in a picture. For audio, it accepts common formats like WAV files and voice recordings, then processes them in short chunks. If a clip is longer than 30 seconds, it's split into smaller pieces at natural pauses. Not every version of Gemma 3 supports both features at once — some handle images only, while the most up-to-date ones handle both. This simply means you need to pick the right model size for the task.

Why should you care? Because it makes free, open-source AI far more useful. Instead of relying on paid services or sending private photos and voice notes to big tech companies, you could run a capable assistant that sees and hears right from your laptop or work computer. Small businesses and hobbyists could automate things like sorting customer images or transcribing short audio clips without monthly subscription fees.

The catch? Running Gemma 3 locally requires a fairly powerful computer. Also, these open models are still not as smart as the biggest commercial AIs at complex reasoning. But for private, everyday help — identifying an object, reading a sign in a photo, or summing up a quick voice note — this is a major step toward AI that works for you, on your terms.