Skip to content

see2023/see_me_now

Repository files navigation

See

Incroduction

  • The purpose of creating this assistant app is to aid children's learning, particularly those with attention deficit.

    • It assists them in solving learning difficulties by providing appropriate reminders when they are distracted and more.
    • Additionally, it can function as a regular AI assistant or ChatGPT client.
  • The app is developed using Flutter and may be suitable for desktop and web platforms, although its primary focus is on mobile platforms (Android/IOS).

    • Please note that Google's ML Kit is only available for Android/iOS.
  • Kindly refer to the usage case below.

Usage

1 Install your own server

Refer to: https://github.com/see2023/see_proxy

2 Install App

  • Compile cp ./lib/data/constants.dart.example ./lib/data/constants.dart
  • Or download the release version

Tips

  • Camera Observer
    • Sitting posture(So far it's mostly head angles) and smile index were obtained from the front camera (no pictures were stored) to judge their current learning status
    • AI Reply
      • Enabled: Send the data above to ChatGPT or a similar interface and let it remind the person in front of the camera.
      • Disabled: Use the fixed strategy to remind the person in front of the camera.
  • Voice Reply
    • Auto Play
      • Enabled: Automatically play the voice reply after the AI reply is completed.
    • Tap to play
      • Enabled: Single tap or double tap the message to play the voice from Azure TTS, and show 3d avatar animation.
  • See Proxy:
    • Your own server address to proxy chatgpt and azure tts
    • Key: Your own key setted in the server(redis)
      • You can share it with your family
      • You can set the usage by key
  • Prompts setting:
    • Select or edit your own prompts in current dialog
    • The first one is used as the background reminder

TODO

  • AI delegated strategy optimization
    • Send tagged data to chatgpt, get response
      • Something like Auto-GPT, but I don’t have the chatgpt4 API permission, and gpt-3.5-turbo just makes me feel annoyed.
  • Fixed strategy
    • Better reminder strategy
      • From camera data
    • Voice observer
      • Listen to his voice and analyze the content and his status
        • The best strategy I've found offline so far is to make him read aloud while he's doing homework
  • Avatar
    • Isolate avatar(mesh) from basic animations
    • Support online avatar link from readyplayer.me or other avatar platform
      • Decorative/Texure configuration
      • Voice match
  • Animation
    • More and better animations
    • Real-time animation
      • From camera data
      • From voice data
  • Better interaction
    • Better voice input
  • IOS and other phones/platforms
    • Tested on Android: Kirin 990, HarmonyOS 2.0(Android 10.0), 8G RAM
  • Reduce token consumption
    • LangChain SummaryMemory?
  • Reduce energy consumption
    • google ml kit usually use cpu as backend...

Thanks

As a back-end developer in a traditional industry, this is my first attempt at developing a mobile app, which also involves a lot of AI and 3D related knowledge. Thanks for the support of New Bing and various projects and organizations below:

DISCUSSION

Please go to discussion

LICENSE

Copy from the components used license