You are browsing as a guest. Sign up (or log in) to start making projects!

ASH

By @Ambush43

  • 1 Devlogs
  • 0 Total hours

Ash is a fully custom, offline-first voice assistant built in Python, running quietly on your laptop. She lives as a small, beautiful floating semi-transparent window with an animated robot face that mixes Visor Bot’s wide eye band, Retro TV scanlines & antennas, and Diamond Bot’s angular corners. The face smoothly changes between six expressive states: idle, wake, listening, thinking, speaking, and sleeping. Wake her hands-free anytime by saying “Hey Ash” — powered by a fully offline wake word detector. She uses Faster-Whisper locally for fast, private speech recognition. For intelligence, she has two brains: Ollama with qwen2.5:7b for complete offline conversations, and Groq’s blazing-fast Llama3-70b for powerful online responses — switch instantly by holding Alt + Copilot key. Ash speaks naturally via Windows TTS and executes real commands: tells time/date, takes screenshots, opens apps/folders, controls volume & brightness, checks battery, locks or shuts down the laptop,

Source
Open comments for this post

@Ambush43 on ASH

22m 15s logged

Devlog #1: Hours of ASH Development 🔥
Date: July 02, 2026
Hey everyone! Today I went all in on ASH — my offline Python voice assistant with the cute animated robot face.
What I accomplished :

Complete project structure setup in C:\ASH
Designed and generated 6 custom animated GIFs for the robot face using a mix of Visor Bot + Retro TV + Diamond Bot style (wide visor eye, CRT scanlines, antennas, diamond cuts). Now Ash has smooth transitions between: idle, wakeword, listening, thinking, speaking, sleeping.
Built the full Tkinter GUI — floating semi-transparent window that automatically moves away when your mouse gets close.
Implemented hotkey system using Copilot key:
Single press → Ash listens
Double press → Minimize to system tray
Alt + Copilot → Switch between Offline and Online mode

Integrated Faster-Whisper (offline STT) + fallback options
Added pyttsx3 + Windows TTS fallback so Ash actually speaks through speakers
Added many real commands: time, screenshot, battery, volume, brightness, open folders/apps, etc.
Added openwakeword support for hands-free “Hey Ash” detection
Fixed major bugs: mic sensitivity, GPU/CPU fallback, Ollama auto-start, window positioning
Set up auto-start with Windows + system tray icon
Cleaned up code, added proper error handling, and optimized for my Lenovo LOQ (Ryzen 7 + RTX 3050A)

Ash is now actually usable! She looks alive, reacts to voice, speaks back, and feels like a real personal assistant.
Challenges Faced:

CUDA DLL errors → forced CPU mode for stability
PATH issues with Python/Ollama → documented full setup
TTS not playing → fixed with fallback
Window too big and annoying → made it small + mouse avoidance

Next Goals:

Add more commands (timer, media control, clipboard, weather, etc.)
Improve wake word accuracy
Make the animation even smoother
Create a proper installer

This was a super productive day. ASH is starting to feel exactly like the loyal robot companion I imagined — always in the corner, calling me “Sire”, and ready to help.
GitHub: https://github.com/AMBUSH-43/ASH
What do you think? Any feature suggestions?

0
0
2

Delete project?

Are you sure you want to permanently delete this project? This action cannot be undone.

All devlogs, followers, and associated data will be removed.

Followers

Loading…