BUILD AN AI SODA MACHINE THAT TURNS ANYTHING YOU SAY INTO A FLAVOR
Say a flavor out loud — even something abstract like the taste of a summer night — and the machine mixes a drink that has never existed before.
by Luke Creatives
Built withRaspberry Pi
- difficulty
- ●●●●●
- time
- eight months
- license
- license not specified
- repo
- repo
●●●●● · eight months · license not specified · repo
WHAT YOU’LL NEED
Jump to section
COMPAREE VERDICT
This is a showcase build, not a how-to. Luke Creatives spent over eight months building a voice-controlled soda machine that uses Google's Gemini to invent flavor recipes from 20 base ingredients, but he published the final video and nothing else — no code, no board files, no parts list, no instructions. Twenty small pumps dose the flavors into a glass, a Raspberry Pi with a touchscreen and microphone runs the show, and the AI lives in the cloud. The hardest part is not the concept, it is the custom pump control board, the plumbing that keeps 20 syrup lines from leaking or clogging, and tuning the AI prompts so Gemini writes recipes that actually taste drinkable. He fried a circuit board and his only Raspberry Pi along the way, which gives you a sense of how far from plug-and-play this is. If you want to build something like it, you are reverse-engineering from a 12-minute YouTube video and starting from scratch on the electronics, the firmware, the prompt engineering, and the flavor selection. The video is beautifully made and the idea is genuinely novel, but this is inspiration, not a weekend project you can follow step-by-step.
IN THE REPO
GOOD TO KNOW
- —No files, code, or build instructions published anywhere.
- —The AI runs in the cloud via Google's Gemini API, so it needs an internet connection and API access.
- —Uses 20 small peristaltic pumps controlled from one custom circuit board — the board design is not shared.
- —Pre-carbonated soda water instead of carbonating on-device.
- —One flavor ingredient was out of stock, so the final machine makes about half a million combinations instead of the theoretical maximum.
- —No licence information because nothing is published.
Parts to buy
11 itemsFrom our check of the build. Exact quantities and part numbers are in the creator’s BOM.
Can I build this?
This build involves mains voltage — for adults comfortable with electrical work only.
Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.
Videos
I built a soda machine that turns whatever I say into a flavor
No build instructions, but shows the machine working and explains the concept.
More builds like this
All projectsGallery
Start here
Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.
- 1.Watch the showcase video to understand what the machine does and how it works conceptually. (This is the only published material. There are no files or instructions to follow.)
- 2.Decide whether to reverse-engineer from the video or design your own version from scratch.(You will need to design the pump control board, write the firmware, and integrate the Gemini API yourself.)
- 3.Source 20 small peristaltic pumps and design a control circuit that can drive them from a Raspberry Pi.(The original used one custom PCB. You could use motor driver boards, but scaling to 20 pumps cleanly is non-trivial.)
- 4.Set up a Google Cloud account and get access to the Gemini API.(The AI runs in the cloud, not locally. You will need API credits and a working internet connection.)
- 5.Choose 20 base flavors and buy the syrups. Test the AI prompt to see if Gemini writes recipes that taste drinkable.(Luke used Claude to help choose flavors. Expect iteration — not every AI-generated recipe will be good.)
KNOWN ISSUES
- There are no files. If you thought this was a documented open-source project, it is not. You are reverse-engineering from a video.
- The custom pump control board is not shared. Designing a clean PCB to drive 20 small pumps from a Pi is a project in itself, and the original builder fried one along the way.
- The AI runs in the cloud via Google's Gemini API, so you need API access, internet connectivity, and ongoing credits. It does not run offline.
- Tuning the AI prompts so Gemini writes recipes that actually taste good is harder than it looks. The video shows successes; it does not show the failures.
- Plumbing 20 syrup lines without leaks, clogs, or cross-contamination is fiddly. Small peristaltic pumps can be inconsistent, and syrups vary in viscosity.
- One flavor ingredient was out of stock during the original build, dropping the total combinations from over a million to about half a million. Flavor syrup availability is not guaranteed.
- Expect months, not weekends. The original build took over eight months and included dead Raspberry Pis and fried boards. This is an advanced, exploratory project.
Are there build files or code I can download?
No. Nothing is published. The only material is the YouTube video.
Does the AI run on the Raspberry Pi?
No. It runs in the cloud via Google's Gemini API. You need an internet connection and API access.
How much does it cost to run the AI per drink?
The video does not say. Gemini API pricing depends on token usage, which varies with how complex the flavor request is. Check Google Cloud pricing for current rates.
Can I use fewer than 20 flavors?
Yes, but you will have fewer combinations. The 20-flavor setup gives over half a million possibilities. Fewer flavors means fewer options for the AI to work with.
Does it carbonate the water itself?
No. It uses pre-carbonated soda water from the store.
Do all the AI-generated flavors taste good?
No. The video shows the successes, but not every combination will be drinkable. Expect iteration and some unpleasant experiments.
Community builds
No community builds yet — be the first, we feature the best ones.
Discussion1
FROM THE COMPAREE TEAM
Over half a million flavor combinations, and the whole thing runs on voice commands and cloud AI — no buttons, no menus. What flavor would you ask it to invent first?
Luke Creatives
Luke Creatives built this AI-powered soda machine over eight months, including a fried circuit board and a dead Raspberry Pi along the way. He used Google's Gemini to generate recipes and Claude to help choose the 20 base flavors. The final video has about 2.7 million views, but no files or instructions were published.
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.
CompareeTEAM1mo ago
Practical notes from our verification: there is no GitHub repository, no published code, no board files, and no parts list. The YouTube video is the only material, and it is a polished showcase, not a tutorial. The builder spent eight months on this and fried hardware along the way, which tells you how far from beginner-friendly it is. If you want to build something like it, you are designing the pump control board from scratch, writing the firmware yourself, and reverse-engineering the AI integration from a 12-minute video. The idea is brilliant and the execution is impressive, but this is not a project you can follow step-by-step — it is inspiration for an advanced build you will have to engineer yourself.