compareeHub

A POCKET AI THAT WORKS WITH NO INTERNET AT ALL

A Raspberry Pi 5, a small screen, two microphones and a speaker turn into a walkie-talkie that thinks — and the entire conversation happens on the device.

by PiSugar

FULL CAD BOM FIRMWARE DOCS

AIOpen-hardware

Built withRaspberry Pi3D printing

difficulty
●●●○○
time
a weekend-plus
license
GPL-3.0
repo
repo ACTIVE526 stars

WHAT YOU’LL NEED

Videos
1
Jump to section

COMPAREE VERDICT

This is a voice assistant that can run fully offline on a Raspberry Pi 5 and fits in your pocket. You press a button, speak a question, and it answers out loud: in the offline build Whisper transcribes your speech, a local model (via Ollama) generates the reply, and Piper reads it back. No internet, no account, no subscription. The repository includes pre-built images, installation scripts, and links to printable enclosures. The Raspberry Pi Zero 2W is supported, but the README recommends the Pi 5 with 8 GB RAM for the offline build. The project also supports cloud APIs (OpenAI, Google Gemini, AWS, others) if you prefer lower latency or smaller hardware, but the offline mode is the compelling part. The single biggest hurdle is configuring environment variables for the services you want (ASR, LLM, TTS): the project includes an interactive wizard (`whisplay configure`) but the first run will still take patience. The enclosures are in a separate repository, so budget time for printing and assembly. If you want a private, portable voice assistant and are comfortable with Linux and npm, this is one of the most complete open projects in its category. If you are new to both Raspberry Pi and language models, expect a steep learning curve at first.

GOOD TO KNOW

  • —The repository includes the full software stack (Node.js, Python, bash scripts) and links to four enclosure designs in a separate repo (github.com/PiSugar/suit-cases), three of them FDM-printable.
  • —The hardware list is clear: Raspberry Pi Zero 2W or Pi 5 (8 GB RAM recommended for offline), PiSugar Whisplay HAT (LCD screen, speaker, microphones), and a PiSugar 3 battery (1200 mAh, or the 5000 mAh Plus version for the Pi 5).
  • —Pre-built disk images are available in the project wiki so you can boot straight into a working system.
  • —The offline build is shown in three video tutorials by Jdaie Lin linked in the README.
  • —Software is GPL-3.0; the hardware design files (for the HAT and battery) are separate products from PiSugar and not part of this repository.
  • —The README is thorough: installation, configuration, updating environment variables, optional wake-word support, image generation, battery level display, and accelerator card integration (LLM8850, Hailo-10H).

Parts to buy

4 items

From our check of the build. Exact quantities and part numbers are in the creator’s BOM.

  • Raspberry Pi 5 (8 GB RAM)Find
  • PiSugar Whisplay HAT (screen, speaker, two microphones)Find
  • PiSugar 3 Plus battery (5000 mAh for the Pi 5)Find
  • MicroSD cardFind

Can I build this?

PrintEnclosure designs in the PiSugar/suit-cases repository (for Pi Zero 2W, Pi 5, and Pi 5 with LLM8850). Print time and filament depend on the case you choose.
BuyRaspberry Pi 5 (8 GB RAM), PiSugar Whisplay HAT (screen, speaker, two microphones), PiSugar 3 Plus battery (5000 mAh for the Pi 5), microSD card
ToolsLinux command line, npm, git, a way to flash the microSD card. Optionally Docker, if you want to run the speech services in containers (the repo has a docker folder).
SkillsComfortable with Raspberry Pi setup, bash scripts, environment variables, and npm. You do not need to write code, but you will edit .env files and run shell commands. If you have never installed software on a Pi, this will be hard.
TimeA weekend-plus: flashing the image and booting, configuring the services, model downloads, and printing and assembling the enclosure
Cost$$: a Raspberry Pi 5 (8 GB), the Whisplay HAT, a PiSugar battery and a microSD card; the software is free
SafetyLithium battery in the PiSugar pack. Standard electronics care applies: do not short, crush or puncture the battery.

Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.

Videos

Offline AI on Raspberry Pi 5 — It Talks, Thinks locally without Wi-Fi! (Complete Tutorial)19:29

Jdaie Lin's step-by-step tutorial (30 Oct 2025): hardware setup, flashing the SD card, Whisplay driver, chatbot install, .env config, Ollama, Whisper and Piper, test run.

More builds like this

All projects
Build a talking AI assistant on a breadboard and host its brain yourselfAIYou can build a 48 GB AI server from scrapped datacenter GPUsAICrypto miners dumped this PS5-based chip, now it runs AI at homeAI

Gallery

Start here

Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.

  1. 1.Download a pre-built image from the project wiki or install Raspberry Pi OS and clone the repository (Pre-built images save time; manual installation gives you more control.)
  2. 2.Install the Whisplay HAT drivers (Audio will not work without the drivers for the HAT.)
  3. 3.Run the dependency installation script and configure environment variables (The interactive wizard (whisplay configure) is the easiest way to set up .env.)
  4. 4.Build the project and start the chatbot service (The README has separate scripts for building, running, and enabling auto-start on boot.)
  5. 5.Print an enclosure from the suit-cases repository and assemble (Four designs are linked: two for the Pi Zero 2W (one of them FDM), one for the Pi 5, and one for the Pi 5 with the LLM8850 accelerator.)

Resources

Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.

KNOWN ISSUES

  • The Raspberry Pi Zero 2W is listed as supported hardware, but the README recommends the Pi 5 with 8 GB RAM for the offline build (Whisper + Ollama + Piper). Do not plan an offline build on the Zero.
  • The Whisplay HAT is a commercial product from PiSugar. You have to buy it.
  • Configuring environment variables is fiddly: you need to choose ASR, LLM, and TTS providers (local or cloud) and set API keys or model paths. The wizard helps, but the first run takes patience.
  • Speech and language models take disk space. Budget time for downloads and use a large enough microSD card.
  • The project supports many cloud APIs (OpenAI, Google, AWS, Anthropic, others) but the offline mode is what makes it distinctive. If you skip the offline setup, you are building a more complicated version of an Alexa.
  • The enclosures are in a separate repository and you have to print them yourself. If you do not have a 3D printer, sort that out before you start.

Can I use a Raspberry Pi Zero 2W for the offline build?

The README lists the Pi Zero 2W as supported hardware but recommends a Raspberry Pi 5 with 8 GB RAM for the offline build (Whisper, Ollama, Piper).

Do I have to use the PiSugar HAT, or can I use a generic USB microphone and speaker?

The project is built around the Whisplay HAT (screen, two microphones, speaker). You could adapt it for generic USB audio but the display, button handling, and audio routing are all written for the HAT. Expect significant rework.

What AI models does it run locally?

In the offline build, Whisper does speech-to-text, a model served by Ollama writes the reply and Piper speaks it. You choose the models in the configuration.

How long does the battery last?

It depends on the Pi model, the PiSugar battery size (1200 mAh, or the 5000 mAh Plus version for the Pi 5), and how often you use the assistant.

Can I use this commercially?

The software is GPL-3.0, which allows commercial use but requires you to release the source of any modified version you distribute under the same licence. The Whisplay HAT is a commercial product sold by PiSugar.

Community builds

No community builds yet — be the first, we feature the best ones.

Discussion1

FROM THE COMPAREE TEAM

The Pi 5 with 8 GB RAM is the recommended platform for the offline build, because Whisper and Ollama are both memory-hungry. What would you rather have: a faster response with cloud APIs or full privacy with local models?

CompareeTEAM6d ago

Practical notes from our verification: the repository is active (last push 28 August 2026) and the README is one of the most complete we have seen for a voice assistant project: installation scripts, an interactive configuration wizard, pre-built images, and links to four enclosure designs. The Whisplay HAT is a commercial product from PiSugar, so you have to buy it. The offline build is shown in three video tutorials, and the project supports AI accelerator cards (LLM8850, Raspberry Pi AI HAT+ 2) if you want faster inference. The single biggest trap is planning the offline stack on a Pi Zero 2W: it is listed as supported hardware, but the README recommends the Pi 5 with 8 GB RAM for offline use. If you are comfortable with Linux and npm, this is one of the most practical offline voice assistant projects available. If you are new to Raspberry Pi, expect a steep learning curve at first.

PiSugar

PiSugar makes battery packs and HATs for Raspberry Pi projects. The Whisplay AI Chatbot is their project for the Whisplay HAT, showing how to build a pocket voice assistant; the offline build is shown in tutorials by Jdaie Lin.

GitHub

Star the project on GitHub

DISCLAIMER

  • Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
  • Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
  • Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
GitHub repo