Google’s best-known AI products, Gemini and the AI features in Search, get most of the attention. Alongside them, the company runs a stream of smaller experiments through Google Labs and Google Arts & Culture that are free to use, require no coding and rarely get much promotion. Five of them stand out for creative work, learning and simply exploring what generative models can do.
All five are labeled experiments, which means features can change or disappear, output quality varies, and availability can depend on your country and account. With that caveat, they are a low-cost way to see where consumer AI tools are heading.
GenType: alphabets from a text prompt
GenType, from Google Labs, generates a complete alphabet from a short description. Type a concept such as letters made of driftwood, flowers, clouds or a peanut butter and jelly sandwich, and the tool produces a set of 26 characters in that style using Google’s image model. The results work as decorative lettering for logos, posters, T-shirt designs or social media graphics. It is not a full typeface tool: punctuation marks are limited, and the output is images rather than an installable font file.
MusicFX DJ: live, generated music
MusicFX DJ is a music experiment developed with Google DeepMind. Rather than producing a finished track, it generates a continuous stream of instrumental music in real time and lets you steer it by typing or mixing prompts such as deep house, ambient piano, lo-fi drums or retro synth. DeepMind describes it as an accessible entry point to live music creation rather than a traditional instrument. It focuses on instruments and texture rather than vocals, so it is better suited to background music, experimentation and finding ideas than to producing a polished song.
National Gallery Mixtape: paintings turned into sound
National Gallery Mixtape is a collaboration between the National Gallery in London and the Google Arts & Culture Lab. You choose paintings from the gallery’s digitized collection, place them in sequence and add stickers that nudge the mood toward a feeling, instrument or genre. The project says Gemini first turns each image into a description, and MusicFX DJ then uses that text to generate the audio, so each sequence of artworks produces its own soundtrack. It is designed as a playful way into the collection, and it also works as ambient music while you work.
Opal: apps without code
Opal, from Google Labs, lets users build small AI-powered web apps by describing them in plain language. Someone might ask for a tool that summarizes YouTube videos or an app that offers one motivational line a day, and Opal assembles a working app as a chain of prompts and model calls that can then be edited visually. Google launched it in the United States in 2025 as a public beta and has since expanded it. The appeal is for teachers, small business owners and students who have ideas for simple tools but no programming background; complex or production-grade apps are beyond its scope.
Talking Tours: an AI guide in Street View
Talking Tours, on Google Arts & Culture, pairs Street View imagery of cultural landmarks with an AI narrator. According to Google, the experiment uses Gemini to analyze the Street View images, together with location names and GPS coordinates, to write a script, which is then voiced by a Google audio model in real time. The result is a virtual guided walk through museums, monuments and historic sites, useful for students, travelers planning a trip or anyone curious about a place they cannot visit.
A glimpse of what is coming
None of these tools are finished products, and that is partly the point. Google uses Labs and Arts & Culture experiments to test ideas in public before deciding what to build into Gemini, Search or Workspace. Features that start as novelties, such as generating music from text or building an app by describing it, tend to show up later in mainstream products in a more polished form.
