In a landscape dominated by closed smart speakers, a game-changing alternative has emerged: an open-source AI assistant designed for those who want to understand and control what's going on under the hood. That gap is filled by Ubo Pod, a device that combines Raspberry Pi 4 or 5, voice, computer vision and sensors., all in a modular, repairable enclosure that doesn't rely on proprietary services to operate.
The proposal goes far beyond asking for music or time. Ubo Pod integrates microphones, speakers, a screen, a camera, sensors, and even an infrared transmitter and receiver. to speak with language models, recognize what it sees, and, in the process, control your TV or any compatible device. You can choose between cloud-based or completely local AI to prioritize privacy, and if you're up for experimentation, it offers a perfect environment to develop your own experiences with voice, vision, and tools.
What is Ubo Pod and what does it aim for?
The developer edition, dubbed Developer Edition (DE), is the heart of the project. Its mission is to replace "black boxes" like the Amazon Echo or Google Nest with open hardware and free software., untethered to a specific platform. Thus, the same team enables speech recognition (speech-to-text), synthesis (text-to-speech), language and vision models (LLMs/VLMs), tool calling, and various activation mechanisms.
Ubo Pod isn't just meant to be hackable; it also wants to be convenient for those who don't want to code. It includes a graphical interface embedded in its small screen and a WebUI to configure it without writing code.This allows you to select AI services, adjust options, and activate features from your device or browser, with a clear and practical approach.
The open philosophy is evident throughout the design. Both the software and hardware are open-source, with public repositories of the Ubo app code and the boards and mechanical files.If you want to audit, modify, or contribute, you have a complete system map, something unusual for this type of consumer assistant.
Design and hardware: modular, expandable and repairable
The Ubo Pod chassis is compact and built to last. It measures 130 x 99 x 52 mm and weighs about 340 grams., with an interior designed for the Raspberry Pi 4 or Raspberry Pi 5 and active thermal management to help maintain stable performance. The modular case makes it easy to replace components, expand your system, and, if necessary, repair it without any hassle.
The front features a 1,54-inch IPS color display with 240 x 240 pixels. Although small, the screen allows for displaying status, menus and an embedded interface to configure and view information on the fly. Next to it, an integrated 5 MP camera provides computer vision. The Ubo Pro 4 variant also supports the use of the official Raspberry Pi Camera Modules V1 and V2; and the Ubo Pro 5 also supports the Raspberry Pi Camera Module 3, expanding the viewing possibilities.
The acoustic section is well equipped for natural conversations. It has a dual stereo microphone with audio capture up to 48 KHz, and stereo speakers with playback speeds of up to 48 kHz and very low distortion (<0,1% THD). The documentation mentions 1 W per channel, while another description mentions 2 W per speaker; in any case, the system is geared toward clear voice and desktop or living-room interaction. For external use, it includes a line output (WM8960) with 40 mW at 16 Ω, jack insertion detection, THD of -75 dB at 20 mW, and SNR of 90 dB with a 16 Ω load.
For physical interaction, the unit includes a keypad with 7 soft-touch silicone buttons and a dedicated power button. It also features an LED ring with 27 addressable RGB diodes (SK6812), compatible with NeoPixel, perfect for visual cues, states or effects.
In terms of connectivity, it inherits the best of the Raspberry. It has Gigabit Ethernet, WiFi 5, Bluetooth 5.0, two USB 3.0 and two USB 2.0 on the SBC, so you won't miss ports to add peripherals or speed up workflows.
One of its differentiating points for the home is its infrared control. It integrates an IR receiver with a range of up to 5 meters (TSOP75238) and four high-power and high-speed 940 nm IR emitters (VSMB10940)., with an omnidirectional pattern. This opens the door to controlling TVs, sound bars, or other devices that accept IR commands, either by forwarding signals from a universal remote or by voice command.
The sensing is also complete. It comes with a PCT2075 temperature sensor (accuracy ±1°C between -25°C and +100°C) and a VEML7700 ambient light sensor., capable of measuring from 0 to 120 kilolux with a fine resolution of 0,0036 lx/ct. This environmental data can be integrated into automation or contextual assistant responses.
In terms of storage, it comes standard with a 32GB microSD card pre-installed with the operating system. The Ubo Pro 5 version adds an M.2 PCIe slot for NVMe SSDs or to add an AI accelerator., an interesting improvement if you plan to run demanding local models or store large databases and memories.
Real Privacy: Physical Curtain and Hardware Microphone Shutter
Beyond discourse, privacy is materialized in physical elements. The camera can be blocked with a physical privacy curtain, so it's completely hidden when you don't want to use it. The microphones, meanwhile, have a hardware kill switch, so you can mute them without relying solely on software.
This commitment is complemented by software architecture. Ubo Pod allows you to work with cloud services or completely locally., maintaining control of your data and adjusting the balance between convenience and privacy as needed. If you want an assistant that doesn't upload audio or images to external servers, you can configure it to do so on your network or on the device itself.
Open source software, no-code WebUI, and over 50 AI providers
Under the hood runs Raspberry Pi OS and the Ubo App, written in Python and published on GitHub. You don't need to buy the device to try the experience: the app can be installed on a Raspberry Pi 4 or 5., so you can experiment with the software before deciding on the full hardware.
AI vendor support is extensive. Ubo Pod supports more than 50 services for speech recognition, speech synthesis, memory, vision, and language models.This list includes cloud options and on-premise solutions, and you can select them from the integrated screen or via the web interface.
If you prefer to choose the models yourself, you can also do so. It is possible to point to API addresses of popular models such as Gemini, ChatGPT or Claude, or use Ollama to run LLMs on the hardware itself, which means low latency and data never leaves the premises.
To develop extensions, there is a gRPC API designed so that you can add capabilities in just a few lines of code. Practical examples include a voice memo assistant (Memo) with memory and context, image generation from voice, image description using VLMs, or controlling the TV with the IR transmitter using voice commands.
The interaction layer is completed with several ways to activate the wizard and with tool calls. Overall, it offers a rich conversational flow that is not limited to text responses., but can execute actions, read sensors, display information on the screen or illuminate the LED ring depending on the status.
Infrared vision and control: your home under your voice
The built-in camera and compatibility with official Raspberry Pi modules make Ubo Pod a device with its own vision. The system can recognize codes, characters, gestures or objects, which opens the door to use cases as varied as reading a code on the screen, detecting a confirmation gesture, or describing a scene.
Universal infrared control is another strong point geared towards everyday life. You can “teach” commands from your favorite remote so the assistant can reproduce them., or ask it to send signals directly to other devices (such as the TV or air conditioning). This is the kind of integration that makes an assistant immediately useful from day one.
Hassle-free user experience… and ready to tinker
Although it is a developer device, it does not force you to be one. It comes functional from the first boot and you can simply configure it from its WebUI without writing a single line of code. From there, you decide whether to stick with the basic experience or dive right in and create your own apps, templates, or integrations.
Those who enjoy lifting the hood will find an ideal playground. Hardware modularity and open software make it easy to replace components and add M.2 storage on the Pro 5. or try AI accelerators. And if you're afraid of breaking something, you can always stick with what comes standard: the customization potential is there whenever you want.
Models, key specifications and differences
The Ubo family is built around the variants that embrace Raspberry Pi 4 or Raspberry Pi 5. Ubo Pro 4 and Ubo Pro 5 share the base (1,54” 240×240 IPS screen, 5 MP camera, dual microphone and dual speaker, IR, sensors, LED ring, 7-button keypad), but they differ on some important points.
In official chambers, Ubo Pro 4 supports Raspberry Pi V1 and V2 modulesWhile Ubo Pro 5 adds support for Camera Module 3The Pro 5 version takes another leap forward with an M.2 PCIe slot for NVMe SSDs or AI accelerators, ideal for fast storage or offloading compute from the processor when running local models.
Both share the SBC connectivity base: Gigabit Ethernet, WiFi 5, Bluetooth 5.0, two USB 3.0 and two USB 2.0. In audio, in addition to the stereo speaker system and microphones, the WM8960 line output with jack detection adds flexibility for external equipment. The package is completed with 32GB microSD card pre-installed with the operating system.
Pricing, availability, and the reality of crowdfunding
The project has reached the community through Kickstarter with a funding goal of $25.000. Rewards start at $109 for Ubo Pro 4 and $129 for Ubo Pro 5It's important to note that neither of these versions includes the Raspberry Pi in the box: you'll need to provide your own SBC. Shipping costs are also not included.
As with any crowdfunding campaign, it's important to be aware of the risks. One of the ads reminds us that “supporting is not the same as buying.”Backing a project doesn't guarantee delivery. Be cautious, review the information, and assess the risk before committing your money.
The official campaign page is: https://www.kickstarter.com/projects/ubopod/ubo-pod-hackable-personal-ai-assistant. The team encourages you to spread the word about the project and join if you are attracted to the technical proposal..
Project status and message from the creator
The Ubo Pod promoter shared that, two days after launching the campaign, he was already approaching 40% of his goal. Ask for a little push to cross the finish line and thank you for spreading the word., especially among technical profiles who are enthusiastic about the concept.
There is one unusual aspect of crowdfunding: A small batch was manufactured in advance and there is some inventory ready to ship immediately to the first backers.That decision reflects the years of work invested in the hardware and software, and the passion behind the project.
The creator's stated goal is clear: Empowering developers to build interactive experiences with voice, vision, and sensors with easeHe also expresses his dissatisfaction with large corporations' devices: closed, tied to services, and, in his opinion, disrespectful of consumer rights.
Regarding visibility, ask for suggestions to amplify your reach. You are familiar with paid marketing agencies, but have decided not to take that route for now., relying instead on community and word of mouth.
Setup and Workflow (without turning it into a tutorial)
Ubo Pod is designed to help you go from zero to useful in no time. The web interface guides the selection of AI providers (voice, text, vision, memory) and basic configuration, while the integrated on-screen GUI allows certain adjustments without having to open the browser.
For those who want to evaluate before investing in the hardware, The Ubo App can be installed on a bare Raspberry Pi 4/5.This way, you can check compatibility, explore the ecosystem, and assess which model is best for you, or whether running it on your own motherboard is enough.
In strict privacy scenarios, You can opt for local or on-premise options, preventing audio and images from leaving your network. And if you're looking for power or specific third-party models, you have cloud providers at your disposal and the ability to target APIs from Gemini, ChatGPT, or Claude, among others.
Case studies and demos now available
Beyond the theory, there are already functional examples that show what it is capable of. Memo, the notes assistant, maintains memory and context in the conversation, making it ideal for capturing reminders or ideas without losing track.
In the visual section, You can generate images from a voice command and obtain image descriptions using vision language models (VLMs). Accuracy and usefulness will depend on the vendors and models you select, but the infrastructure is in place.
For the living room, IR control shines brightly. It is perfectly feasible to tell the assistant to change the channel, adjust the volume or turn off the TV., whenever the device receives infrared commands. Combined with environmental sensing (temperature and light), you can create simple contextual behaviors.
Why choose an open assistant
Closed assistants offer convenience, but at the cost of control and flexibility. With Ubo Pod, you choose providers, decide where your data is processed, and can inspect the code.For many, this isn't a detail, but rather the difference between using a tool and being "used" by it.
Additionally, the hardware is not soldered onto a proprietary board from a single vendor. Using Raspberry Pi allows you to upgrade, repair or reuse components, something especially valuable in a context of rapid technological obsolescence.
Limitations to consider
Like any open hardware project, there are important nuances. Rewards do not include the Raspberry Pi, so you'll have to factor in that cost and availability. In audio, according to the documentation, you'll see figures mentioning 1 W per channel and also 2 W per speaker; if maximum volume is crucial for you, consider the context of use (desktop or small room) or using the line output with external speakers.
On the other hand, it is an evolving project. The list of AI providers, features, and local performance depend on the community, dependencies, and the Pi you build (4 or 5).The good news is that by being open, you'll be able to adapt and improve over time.
Ubo Pod fits into a niche that increasingly demands more voice and more control: people who want a useful assistant, with vision and sensors, but without giving up privacy and the ability to dig around whenever they want. With modular hardware, open software, support for over 50 AI providers, local options, and a no-code WebUIThe project puts a real alternative to the black boxes on the market in the hands of the community. If you're interested in the idea, the Kickstarter campaign—with prices starting at $109 for the Pro 4 and $129 for the Pro 5, not including the Raspberry Pi or shipping—is the starting point for supporting and, if you're among the first, even receiving units from the initial batch already manufactured.