top of page
Search

Splash Music AI Generator Shuts Down, Pivots Back Into Video Games

Splash Music was an AI text-to-song web app, featuring a generative music engine with AI vocals. After eight months in business, the company announced in March 2024 that they would be discontinuing the service and returning to their roots in game development.


The company's CTO, Lex Toumbourou, confirmed that their existing Roblox game continues to draw millions in monthly active users. Splash also has a new title on Steam that's scheduled for a tentative future release.



In this article we provide a general overview of how the app worked, including its core features and comparisons to other AI song generators like Suno, Riffusion and VoiceMod.


Table of Contents


How did Splash Music wok?


The text-prompt interface was a core feature at Splash. Users described music and could quickly generate song cues in that style. A few musical parameters like BPM, key signature, and scales (modes) were available for those who needed it.


Text prompt field in Splash Music

Within 20-30 seconds, the system generated four short samples accompanied by AI-generated images. Each sample could be auditioned with playback controls and expanded via the select button.


There was a dropdown arrow used to generate similar variations on the reference track. If none of the music felt right, Splash also offered a more unique samples button that created entirely new music. Here's what that interface looked like:


Splash music dashboard

One of the best features at Splash was the ability to store each new step in your generative process. Prior choices were kept on-page while the system introduced newly generated content. The layout was comparable to AI chat apps like ChatGPT and Perplexity. It was a big improvement on the AI music interfaces from Google and Meta.


Generating music in Splash

Once a sample was selected, Splash could expand it into a 15 second clip, as shown above. A new set of options became available. Users marked their favorite tracks, downloaded and shared the audio files and then used the Edit Song button to make changes to the duration. Paying subscribers had the option to generate songs that were 30 seconds or longer.


Choosing a song duration

The edit song panel also featured a Vocals section. This gave used the option to add lyrics to a song. In the past, most AI vocal generators ran separately from an AI music generator. The AI rappers at Splash were a big part of what gave the product character, differentiating it from text-to-music services like MusicGen, MusicLM, and Stable Audio. Listen to a side-by-side comparison of Splash to these other AI music generators here.


How do you add AI vocals to songs on Splash?


To add vocals to Splash songs, users began by clicking the add lyrics button. A new text area appeared with several options for customization. Users could write their own lyrics into the box or use the prompt magic tool labeled Lyrics to get help from an AI lyric generator.


Splash music lyric interface

After clicking the lyric button, a popover asked for a short prompt that indicated what the song should be about. It could generate simple four-line stanzas or create enough content for a full length song. The create lyrics button closed the window and populated the text area in your song dashboard.


AI lyric generator

Next, you needed to select an AI voice to perform the lyrics. The free version of Splash came with only one option; an AI rapper named Jacob.


A total of three rappers and ten singers were available on the paid plan. Each vocalist had a different timbre and cadence. Clicking the play button next to a name let you hear what they sounded like before committing to the paid plan.


Selecting an AI vocalist

To hear more than one AI voice on a track, users could add lyrics a second time and type in another verse or chorus. Clicking on a new AI voice name and selecting someone else instead produced new segments with that voice. The song-section modules could be dragged up and down on the vertical timeline, to hear how each part sounded in a different order.


Finished AI song example

The library dashboard (shown below) include the options to view all of your past generations or highlight only your favorite ones:


Splash music user library


What did AI songs from Splash sound like?


Splash supported a diverse range of music genres, but their sound design was exclusively electronic. On the one hand, this meant that users could enjoy high fidelity audio and music that kept a consistent tempo. However, it also made it hard for Splash to compete with realistic sounding AI song generators like Suno.


Watch the video below to listen to some examples of the kind of music and album art that Splash created:



While they were active, Splash partnered with influencers like Amli and companies like the AI Music Video Generator Neural Frames, creating trippy visuals that go will with AI generated music.


An example of the Neural Frames collaboration can be seen below, where lofi beats were coupled with colorful, animated sea creatures:



Alternatives to Splash (Suno, Riffusion, Voicemod)


The best alternative to Splash, now that it's been discontinued, is the text-to-song app Suno. Compared to Splash, Suno sounds incredibly realistic and offers a generous free plan. The second best option is Riffusion. Both offer similar lyric-to-vocal workflows with AI generated album art, just like Splash.


A third option is VoiceMod's text-to-song app, also called a 'meme machine'. Despite the silly name, Voicemod's app is an impressive piece of tech. It lets users select from preset genres, choose from different AI vocalists, and enter lyrics to be turned into vocal melodies on the track.



Voicemod does not generate AI music from scratch. As a meme machine, the company's goal is simply to make funny content that users can share with friends. If that's your goal, the app is free to use and you can scratch the creative itch without paying anything.


Comparing Splash Music to AudioCipher


AudioCipher is another generative text-to-music software company, but it caters to a different crowd. It's a DAW plugin and standalone app that lets users transform words and phrases into melodies or chord progressions.


AudioCipher v3

Both AudioCipher and Voicemod support the option to pick key signatures and control the music's rhythm, but AudioCipher exports in MIDI format instead of mp3 or wav. It's a creativity and inspiration tool for musicians rather than a "finished song" generator. You can watch demo of the app below:



What are Splash Music's Terms of Use for publishing?


The Splash Terms of Use indicated that music created on their platform could be published under a royalty-free license.


However, they reserved the right to "limit your use of the Recordings or require you to cease using them (and delete any copies of them) if we form the view that your use of the Recordings would infringe the rights of any third party, including Open AI."


Their terms further indicated that "You grant us a royalty free, irrevocable licence to reproduce, communicate and otherwise use and exploit any Song Prompts or User-Submitted Lyrics submitted by you, to the extent required for us to operate Splash AI."


This meant that Splash Music users enjoyed only limited control and ownership over the lyrics and songs they created.


Nevertheless, the app was loved by many and a general sentiment of disappointment was expressed on social media when the news was announced.


We look forward to seeing how they incorporate their AI music engine into future gaming experiences like the one currently listed on Steam.

Comments


bottom of page