PoLang
Zero photo uploads. Your privacy stays on-device. Search your local gallery and edit photos β all in natural language.
β Free
β No ads
β On-device privacy
Zero photo uploads. Your privacy stays on-device. Search your local gallery and edit photos β all in natural language.
A smart gallery that speaks your language
Manage your gallery by chatting: "Find last summer's photos" or "Make this one warmer." PoLang understands and does it.
Auto-tags every photo (scene / people / objects / activity), with natural-language search, smart albums, and AI collections.
GPU-accelerated pro tools: smooth skin, brighten, slim face, enlarge eyes, lip color, blush. One-tap AI enhancement.
Leica (classic / vivid / mono), film (gold / Fuji / vintage), real-time GPU effects, with batch support.
All AI and face processing happens locally. Your photos are never uploaded, and we collect no data.
Browse, search, and edit your gallery from Feishu or Telegram β results land right back in the chat.
A quick look
Make the AI truly get you
Tagging, face recognition, and semantic search all rely on on-device AI models. Download them in "Model Center β Required" β 7 models, ~1.5 GB.
Open "TAG Generation" and scan so PoLang learns the people and scenes in your gallery (faces + semantics), about 1.5h / 10k photos. Remember to keep background activity on.
Want "beach" or "night scene" searches to be sharper? Run image understanding to tag scenes, objects, and activities β about 24h / 10k photos, best overnight while charging.
AI chat runs on a remote model by default. Sign in with email for a higher free quota (guests get less), or switch to the local model to chat offline.
Name people and set relationships (family / friendsβ¦), then add facts in "AI Memory" β PoLang gets to know you better over time.
Connect Feishu / Telegram to drive your gallery from IM in natural language β results come right back to the chat. Telegram uses long-polling, no public IP needed.
"Find last summer's beach photos" β matches show up, grouped smartly. All search, face, and scene recognition happens on-device; photos are never uploaded.
Drive your gallery with natural language
"Find last summer's beach photos"
ποΈ 12 matches, grouped smartly by time
"Make this one warmer"
π¨ Warm filter applied β what you see is what you get
"Smooth skin 50, brighten 30"
β¨ Beauty params applied
"What's in this photo?"
π€ Beach, sunset, two people (recognized on-device)
No. All AI and face processing runs on your device. Your photos never leave your phone.
Smart search relies on on-device models and a one-time gallery scan. First download the required models in "Model Center" and run a scan β see the Getting Started guide.
The required face + semantic scan takes about 1.5h per 10k photos; the more accurate image-understanding pass takes about 24h per 10k β best run overnight while charging. Already-processed photos won't repeat.
Chat defaults to a remote model with a free quota; sign in for a higher quota, or switch to the local model to chat entirely offline.
Yes. No ads, no subscriptions, no in-app purchases β every feature is free.
The on-device AI (Qwen) works offline. The remote AI (DeepSeek) is optional and requires a network.
Android 7.0 (API 24) and above. Beauty and on-device inference run best on arm64 devices.
No. We collect no user data. See the Privacy Policy.