DIY AI Can Be Beautiful
Jdaie Lin's offline voice assistant powered by a Raspberry Pi 5 and a Whisplay HAT just got a slick case and some new features.
At this stage of the game, personal AI assistants made by hobbyists are better than anything you can buy commercially. But there is one way that these DIY devices almost always fall short — their appearance. They may be loaded with cutting edge features, but nine times out of ten, they consist of bare circuit boards, a mess of wires, and maybe a shoddy 3D print on some of the better creations.
Consider the offline AI assistant recently created by maker Jdaie Lin. This Raspberry Pi 5-powered device uses a Whisplay HAT to simplify the development of a polished, portable voice assistant. And with a Qwen3-1.7B large language model running locally, it is also quite useful. However, this device is just a bunch of circuit boards stacked on top of each other. Good luck getting that through airport security or avoiding a short circuit if you set it on a conductive surface.
To right this wrong and complete the device, Lin just upgraded the voice assistant. Not only was the appearance transformed with a new case, but some additional features were added in as well.
For starters, Lin created a design and 3D printed a case. It exposes only the display, covering all of the other internal components. Separate, springy pieces were made for covering the buttons, so even they got a makeover. The final result looks quite polished, but exactly how good a copy might look depends on the quality of the printer being used.
To support new features, a Raspberry Pi Camera Module 3 was added to the device. The onboard language model can be used to answer questions about the pictures it takes — although this takes about two minutes of processing time on the Raspberry Pi. Lin points out that this can be sped up to just a few seconds by offloading the work to a nearby computer running an AI model with Ollama.
Those that don’t mind going online can unlock more features yet. Using OpenAI’s API, image generation capabilities can be added to the device. Lin also showed how photos captured by the onboard camera can be edited with voice commands. The results are shown right on the device’s own display.
If you already have built a copy of this voice assistant, then you have everything you need to finish it off. Lin has made the design files freely available to anyone that wants them.