Installing Ollama for AI Features
Fictura's automatic photo/video title, description, and tag generation during import runs on Ollama, a local AI model runner — not a cloud service. If Ollama isn't installed and running, that specific feature is simply unavailable; nothing else in the app is affected, and you can start using Fictura fully without it.
What actually needs Ollama
Only vision-based metadata generation depends on Ollama — the Vision status shown in the Import view's AI Options. Two other AI features are easy to confuse with it but use different, separately-installed tools:
- Audio/video transcription uses faster-whisper (or a compatible
whisperinstall), not Ollama — see the Import view's Audio status indicator, separate from Vision. - Face detection uses a locally bundled model, not Ollama at all.
So if you only care about transcription or face detection, you can skip this guide entirely.
Installing and running Ollama
Follow the official install guide at ollama.com's quickstart guide — it covers macOS, Windows, and Linux, and leaves you with Ollama installed and running in the background. Fictura talks to it over http://localhost:11434 by default, the same address you get by installing Ollama with its defaults untouched.
Pulling the vision model
Fictura is configured out of the box to request llava:34b for image analysis. Pull it once before your first AI-assisted import:
ollama pull llava:34b
This downloads several gigabytes and only needs to happen once per machine. If you've changed the Vision Model field in Settings under Import's AI features to a different model, pull that model's name instead — whatever's configured there is what Fictura actually requests.
Confirming it's working
Two ways to check:
- From a terminal:
ollama listshows every model you've pulled;curl http://localhost:11434responds if the service is up. - From inside Fictura: open the Import view, expand AI Options, and look at the Vision status indicator next to the "Generate Metadata" checkbox — green means Fictura can reach Ollama and use it right now. If you just started Ollama and it still shows unavailable, click the refresh (↻) button next to the status indicators to re-check without restarting the app.
If Vision stays unavailable, double-check the base URL Ollama is actually listening on matches what's configured in Settings (Import → AI features → Base URL) — this only needs attention if you've moved Ollama off its default port or host.
What happens if it's not running
Nothing breaks. The "Generate Metadata" checkbox in Import's AI Options is simply disabled when neither Vision nor Audio is available, and import proceeds exactly as normal — files still get imported, EXIF extracted, thumbnails and previews generated, and a baseline version created. You just won't get automatic titles, descriptions, or tags for that import; you can always add those by hand afterward, or come back and re-run Sync's AI-assisted options once Ollama is up.