# A 1984 Omnibot gets a Pi 5 brain and a Ring-O eye

A battery-eaten Tomy Omnibot from 1984 now drives itself from a Raspberry Pi 5, steered by a cassette-tone command language and a companion robot that tracks you with one cyan eye.

- Category: Hardware (mod)
- Device: [Tomy Omnibot](https://afterware.xyz/d/tomy-omnibot)
- Made by: [Mario Cruz](https://mariothemaker.com/) on Make:
- Original: https://github.com/MarioCruz/omnibotAi
- On Afterware: https://afterware.xyz/p/a-1984-omnibot-gets-a-pi-5-brain-and-a-ring-o-eye

## Can I build this?

- Difficulty: Advanced
- AI-buildability: 3/5 (The tone synthesis, YOLO tracking and steering rules are software AI can write, but restoring a corroded 40-year-old board, repairing the wheels and integrating the hardware is hands-on work.)
- Time: A few weeks
- Cost: $250–450
- Parts: Raspberry Pi 5, IMX500 AI camera, Bluetooth cassette adaptor, OLED display, Tomy Omnibot
- Tools: soldering iron, isopropyl alcohol, 3D printer

Estimates based on the original write-up. How the AI-buildability score works: https://afterware.xyz/ai-buildability

## Build sheet

- Device: Tomy Omnibot (1984)
- Brain: Raspberry Pi 5
- Vision: AI camera running YOLO
- Control: Cassette tone commands

## Starter prompt

```
Based on "A 1984 Omnibot gets a Pi 5 brain and a Ring-O eye" on Afterware: https://afterware.xyz/p/a-1984-omnibot-gets-a-pi-5-brain-and-a-ring-o-eye

Write a Python script for a Raspberry Pi 5 with an AI camera running YOLO detections. Compute the horizontal center of the largest person bounding box, then output "left", "forward", "right" or "stop" (stop when the box covers more than 60% of the frame). Send the command over a serial or audio link to a robot base.
```

## Story

Tomy's 1984 Omnibot came to Moonlighter FabLab with battery acid eating its main board. Mario Cruz spent hours with isopropyl alcohol, a toothbrush and a soldering iron to bring its 40-year-old logic back. Ten seconds into the first drive, the rubber wheels fell apart. The restoration kept growing from there.

The hook was the original cassette deck. The robot read audio tones as motor commands: 1614Hz for forward, 2208Hz for left, 1422Hz to switch the cassette audio to the speaker. A Bluetooth cassette adaptor and a Raspberry Pi script that synthesized those tones let a laptop drive the robot across the kitchen. A spare AI camera running YOLO, plus a companion robot named Ring-O built by Placitech, turned it into something that tracks people.

The lesson is in the controls. An LLM answered "right" for every frame position, so Cruz replaced 200 lines of prompt engineering with about fifty lines of arithmetic: left of center turns left, centered goes forward, more than 60% of the frame means stop. Anyone with an old robot and a camera can steal that rule set.

## Resources

- [Project code](https://github.com/MarioCruz/omnibotAi)
- [Creator site](https://mariothemaker.com/)

Tags: omnibot, raspberry pi 5, yolo, vintage robot, upcycled, computer vision, 3d-printed, vibe-coded

AI assistants can search all Afterware builds through the MCP server at https://afterware.xyz/mcp (setup: https://afterware.xyz/agents).