Key Takeaways
- Google’s experimental MusicFX and ImageFX tools are generative AI mechanisms that use keywords to assist users.
- MusicFX can generate 70-second songs with a variety of prompts and styles.
- ImageFX has improved its text prompts for image generation and, despite some odd results, shows potential.
Generative AI It has surprised and dismayed people with its ability to convert text prompts into images and then back to text. But earlier this year, Google starts testing AI They are designed to generate something even if you don’t have any ideas to type into the prompt box. As part of Google Test Kitchen, ImageFX and MusicFX help suggest what users should ask to generate wacky ideas for images and music. music .
Related
11 tedious tasks that Google Gemini handles for you
Among other important updates at Google I/O, the Gemini 1.5 Pro will soon be able to answer questions about the world around you using video.
The technology itself isn’t all that far removed from Google’s more widely known Gemini — in fact, ImageFX uses the same text-to-image diffusion model as Gemini and Google Lens — but the experimental program aims to spark more ideas and creative thinking by providing a list of alternative keywords to use in a prompt.
We had Google’s MusicFX create some songs, and ImageFX even made an album cover and band poster. But did Google Test Kitchen leave a bad impression, or is this tool the future of AI?
Related
5 cool things Google’s Gemini AI can do for the Pixel 9
The new Google Pixel 9 smartphone comes with some unique AI features.
What are MusicFX and ImageFX?
An experiment or something more?
MusicFX and ImageFX are experimental AIs that Google is currently testing, along with similar options like TextFX. Google’s FX tool is a generative AI to use when you don’t know how to write a prompt. This web-based software is designed for experimentation and exploration, rather than the productivity focus of, say, Gemini in Gmail. Both can be tried for free on the Google Test Kitchen in the US.
MusicFX turns text prompts into short songs up to 70 seconds long. This experimental AI helps users write the prompt and suggests what to say before they type anything into the box. As you add a prompt, the software turns keywords into drop-down menus called Chips. For example, the prompt “Play Country Music Inspired by Crows on Guitar” offered multiple suggestions for country, crow, and guitar all together. With just a few clicks, I was able to change the original prompt to “Play Blues Music Inspired by Owls on Harmonica.”
Related
10 Gemini Live features you’ll want to try
Google’s AI sounds more human-like, but what exactly can Gemini Live do?
Similarly, ImageFX is an AI image generator that helps improve text-based prompts. It’s a large-scale language model powered by Imagen 2, the same technology that Gemini uses. While Gemini can already generate images, ImageFX suggests modifications to the prompt, turning keywords and phrases into drop-down menus. These so-called chips are designed to give users more ideas and make the results even more refined.
ImageFX doesn’t even require an initial prompt.
In fact, ImageFX doesn’t even require an initial prompt: simply type “I’m feeling lucky” and a random prompt will be generated. “Dreamy pastel landscapes, soft lines, gentle colors, fluffy clouds, iridescent mountains, minimal” can turn into “mystical neon portraits, angular shapes, bold colors, dramatic clouds, jagged mountains, decorative,” and you don’t even have to tap on the keyboard.
Related
Gemini Live is here, allowing you to have voice conversations with Google’s AI
If you have Gemini Advanced, you can get started right away.
A short jingle generated by MusicFX
Soulless, but not terrible
When I first tested MusicFX, I quickly learned what the music generator could and couldn’t do. Its features were both a relief and a disappointment. First, I couldn’t get MusicFX to generate vocals, but it did generate some Do-Re-Mi vocals when I requested an acapella. And, much to the relief of artists, the program can’t recreate a specific artist. Sorry, but MusicFX won’t be churning out Taylor Swift songs anytime soon.
MusicFX creations are currently limited to 70 seconds long, but if you turn on the loop option they’ll play seamlessly. Songs are 30 seconds long by default, but you can adjust the length by opening the settings menu.
Related
SearchGPT Explained: What is SearchGPT and How to Try It First
There have been rumors for some time that OpenAI is working on a competitor to Google Search, but now it’s finally happening.
Prepared to hear music that was just as bad as the three-finger melted face photos of my early image generators, I was surprised when I clicked play and found that the song wasn’t terrible. It was the kind of music you’d imagine playing in an elevator, or in the background while on hold. Since the initial results weren’t too bad, I generated a few more, trying multiple genres, speeds, and instruments.
The music lacks the soul and emotion of songs that I would normally find myself singing along to.
After a while, the songs the software generated all started to sound similar (in retrospect, maybe I shouldn’t have asked for so many country songs). The clips were short, but without any sense of structure like a chorus or verse, it seemed like a short, repeated beat with subtle variations. I didn’t shiver, but I didn’t hum or tap along to the beat either. The music lacked the soul and emotion of songs that I would normally find myself singing along to.
Related
ChatGPT free users can now generate DALL-E 3 images, but limited to two per day.
ChatGPT now allows you to generate images from text without a subscription.
At times, the software struggled to hear precise instructions — for example, when I requested music to be played only on an acoustic guitar, it produced a song featuring multiple instruments. While I can’t imagine MusicFX creating Billboard hits, I can imagine it generating background music for video ads and commercials. But copyright disputes surrounding generative AI mean it’s unclear whether the resulting images could or should be used commercially.
One of the best features of AI is its outlandish randomness, which can sometimes lead to great ideas.
But the best thing about MusicFX is the drop-down chip designed to bring out more ideas. In my opinion, one of the best features of AI is the outlandish randomness that sometimes results in great ideas. It’s fun to use the different suggestions to try something new or take one idea further. The way it comes up with outlandish ideas like “cheerful, optimistic cyber pizza party music in an underwater arcade” is a lot of fun to experiment with, but I was disappointed when I Googled that suggestion prompt and it spit out the same suggestions many times before.
However, the best way to customize the result is to use DJ Mode. This option provides sliders for each part of the prompt, so you can increase the tempo for an upbeat or lower it for a classic rock vibe. This method gives you more control over the end result. As ideas come to you, you can add to the list or use the suggestions at the bottom. However, DJ Mode still lacks download and share functionality.
I used MusicFX to generate country songs, psychedelic jingles, and, in DJ mode, classic rock with acoustic instruments.
Related
Does Apple Intelligence actually have a shot at winning the AI race?
If Apple can beef up its in-app tools, it has a chance to stand out from the competition.
Some of the ImageFX results were scary, but some were impressive.
The software suggested quick adjustments to reorient the photo.
Google / Pocket-lint
Naturally, after coming up with a few different AI-generated tracks, I needed to create album covers to go with them. For that, I used ImageFX, a tool powered by the same Imagen 2 as a subset of Gemini that generates graphics. Like MusicFX, it uses a chip to suggest tweaks to your prompts, from style to what gets generated.
The AI achieved exactly the style I was going for.
My first prompt was for a clown-like musician with three arms and white eyes, the character that will probably haunt my nightmares now. Remembering how hard it is for an AI to recreate the human form, I tweaked the prompt and was surprised at how quickly I found something I liked. The AI perfectly replicated the style I was going for, reminiscent of vintage circus posters.
But what was most impressive was that the AI was able to process text. Previously, the AI I’d used would fail to add words correctly, even when I simply asked for “Happy Birthday,” and would generate gibberish or misspelled words. But when I told ImageFX what words to add, it added the correctly spelled words. It’s not perfect — when I didn’t tell it which words to add to my album cover, it added letter-like shapes to parts of the design that were clearly intended for text — but it’s more impressive than the text in the images ChatGPT tried to generate.
Related
Tim Cook reveals when ChatGPT will be added to iOS 18
During Apple’s latest earnings call, the CEO confirmed that ChatGPT integration is coming soon.
Here are some images generated.
Are MusicFX and ImageFX in Gemini’s future?
If there’s one feature I’d like to see in Gemini, it’s chips.
Generative technology, especially technology that attempts to replicate art, makes us wonder what place it has in our future and how it will impact actual human creativity. If MusicFX is any indication, AI-generated songs will become our on-hold music, elevator music, or forgettable background music for social media videos. I can’t imagine listening to any of the songs this tool has created so far at full volume in my car. But for now, MusicFX is experimental and could move forward in a big way.
Another problem that any machine learning platform must solve is where the training data comes from. Google hasn’t revealed where it finds the music to train its system, although a Billboard report said the company used copyrighted music in its training set. Legislation could play a big role in whether MusicFX makes it out of Google Test Kitchen, as litigation continues over the use of copyrighted images in training data.
Related
Gemini and Google Workspace help you be more productive in most cases.
Google’s Gemini is great for summarizing Google Docs and emails, but it’s a bit odd when it comes to spreadsheets and other Workspace tools.
