Skip to content

Openflow 2.0: Pixel Pet Companion, Sound Feedback, GPU Acceleration & Local AI

Latest

Choose a tag to compare

@SadmanMahi67 SadmanMahi67 released this 24 May 16:36

This update transforms Openflow with a pixel pet companion that shows your AI status through animated GIFs, configurable sound feedback, GPU acceleration, custom Whisper models, and offline local AI with llama.cpp.

Pixel Pet System

  • A companion pet overlay that shows real-time AI status (idle, running, waiting, jumping) via animated GIFs
  • Import custom pets from ZIP files or browse pets at codex-pets.net
  • Built-in YoRHa Sit 2B default pet
  • Native OS window dragging with -webkit-app-region: drag
  • Click the pet to trigger a waving animation

Configurable Sound Feedback

  • 7 synthesized capture start sounds (Pop, Click, Rising, Blip, Tap, Whoosh, Pulse)
  • 7 synthesized paste done sounds (Chime, Ding, Ascend, Descend, Sparkle, Chord, Confirm)
  • Preview any sound directly in Settings before selecting
  • Zero dependencies — sounds are generated by your sound card

GPU Acceleration

  • A newly bundled Vulkan runtime allows compatible PCs to use the GPU for processing transcriptions

Custom Whisper Models

  • Download and switch between Small, Medium, Large, and Large Turbo models directly from the app

Offline Local AI

  • Integrated llama.cpp support — download the runtime and models to run text refinement completely offline
  • Supported models: Llama 3.2 3B Instruct, Gemma 3 4B IT, Qwen 2.5 3B Instruct

Visual Polish

  • New default Plasma shader background with darkened preset colors
  • Compact glass-morphism card design across all settings tabs
  • Fixed color picker race condition

Other Features

  • Custom Prompt Filters — create and apply custom filters to fine-tune transcriptions
  • Global hold-to-talk hotkey: Ctrl + Win
  • Refinement styles: Casual, Formal, Summarised, Bullet Points, Email Ready, None
  • Auto-paste into the active window
  • Vocabulary list for names, brands, and domain terms
  • Transcription history with raw and refined text