⚙ Settings
Configure Pictaria — changes apply immediately, no restart needed. Only the port, data paths, password, and backup location live outside this page, in the server environment.
Connect with the app password to edit settings.
▶Server
How Pictaria reaches your Immich library. Values marked env come from the server environment until you override them here. Changes apply immediately — no restart.
▶AI Providers
Set up only the providers you plan to use. Feature-specific provider choices remain in Enrich, Curate, and Voice TTS.
NOTE: Enrich requires vision models with structured output and the Curate referee needs multi-image support.
▶Insights
How often the snapshot refreshes, and what counts as a trip on the timeline. Trip thresholds apply at the next refresh. Details in the Insights guide.
▶Enrich
Controls enrichment, the image size sent to AI, and the prompts and taxonomy that govern what the model may say. Choose the provider for new runs on the Enrich page; your selection is remembered. Provider connections and model names live under AI Providers. How the pieces fit — prompts, taxonomy, and the response schema — is explained in the enrichment guide.
▶Discarded Photos
Photos discarded from the Enrich stuck-photos list. Nothing was changed in Immich.
Restore a photo and enrichment will try again.
▶Curate
How the review queue presents your photos. Changes apply on the next page load — decisions you already made are never affected.
▶Devices
Every device running Pictaria Frame this server has seen — the device names come from the Pictaria Frame setup screen.
Deleting a device removes it from the Remote and erases its display history; if it comes back
online later it simply registers again. Use it to clean up renamed, re-onboarded, or retired frames.
▶Location Groups
Group nearby cities into one synthetic location. Useful for handling neighborhoods or suburbs — e.g. a "Bay Area" that includes San Francisco, Burlingame and Oakland.
Once created, Insights shows the group (marked ⁕) instead of its member cities, and also lets you easily create Albums for the group.
NOTE: Groups only relabel in Pictaria — your photos in Immich are untouched, and deleting a group brings the original cities back.
i
▶Wake Words
Upload an openWakeWord TensorFlow Lite model for Pictaria Frame devices with the one-time Pictaria Connect in-app upgrade. Models stay on this server,
download to a frame only when selected, and continue working offline after installation. Pictaria checks
the model structure here and the Android runtime validates and smoke-tests it again before activation.
▶Voice TTS
unsaved changes
Configure how Pictaria answers spoken questions and turns replies into speech. Provider connections live under AI Providers; Voice Commands and the wake phrase are configured in Pictaria Frame. Changes apply immediately.
Interesting and Tell Me commands
Models used to reply to “interesting” and “tell me …” voice commands. Choose a fast provider so Frame can answer promptly. When invoked, “Interesting” sends the current photo’s Immich preview plus its date, location, and filename to this provider, independently of the Enrich switch. It requires a vision model. Selections use the connection info under AI Providers.
Uses the connection configured under AI Providers.
Vision model for photo questions. Leave empty to use the provider’s default. LM Studio supports this command only when Pictaria Server runs on macOS.
Text model for brief answers. Leave empty to use the provider’s default; a small, fast model is usually best.
Maximum time for the full answer, including Immich lookups. Frame stops waiting after 45 seconds, so the server limit is 40 seconds.
Text to speech
Turns Pictaria’s answer into audio for Frame. Choose None to disable spoken replies.
Unsaved changes
Try text to speech
Generates and plays a test reply through the server.
Try a voice command
Routes a phrase exactly as Frame would. Tests here don’t count toward Frame metrics.
▶Prompts
The exact instructions sent to the AI provider for “interesting” and “tell me…” voice commands. Leave a box empty to use the built-in prompt, or load it, edit what the answers should emphasize, and save. Answers from a custom prompt carry a “-custom” promptVersion so odd results are traceable. Clear a box and save to return to the built-in prompt.
▶Location Names
How photo GPS becomes place names — the labels on the frame, voice answers like “where was this taken?”, and Insights' home-base name. Optional place naming sends coordinates to the Geoapify account you configure. Weather sends the frame’s city or US ZIP to Open-Meteo for geocoding, then requests the forecast with the resulting coordinates; no key is required.
▶Backups
Automatic snapshots of Pictaria's own data — your Curate decisions, tags and captions, album rules, location groups, and these settings. Your photos live in Immich and are not part of this. The destination folder is set by
BACKUP_DIR in the server environment — point it at a NAS mount or synced folder for real safety. See the backup guide.▶Appearance
Light mode
Use a white background instead of the dark, photo-first theme. Applies to this browser on every Pictaria page.
▶Support Pictaria
Pictaria Server has no telemetry, analytics, or Pictaria-operated cloud service. Features you enable can connect to your Immich server and the AI, TTS, weather, or geocoding providers you choose; see the privacy overview. If Pictaria has earned a place in your home, you can support its development with a one-time contribution at pictaria.ai/contribute — you'll receive a supporter key by email. Paste it here for a small badge and our sincere thanks. Supporting never gates any feature. The key is verified entirely on this server against a pinned public key; verification makes no network request to Pictaria.
Supporter key