You can get the instrumental from a song by uploading your MP3 or WAV file to an AI stem splitter, which separates the vocals from the music tracks.
When you want to separate the music from the vocals, you can use an AI tool to process the audio. People often look at sites like moises.ai, lalal.ai, manual.audacityteam.org, help.moises.ai, isolate.audio, and podcast.adobe.com to see how to do this. On Stem Split AI, you can make yours by uploading an audio file directly to the site. The software uses Demucs htdemucs, an open-source model developed by Meta AI research. We run this model on our own servers to process your audio. When you upload a track, the model analyzes the audio and divides it into five distinct parts, known as stems. These five stems are the vocals, the instrumental, the drums, the bass, and a category for other sounds. The instrumental stem is created by adding the drums, the bass, and the other sounds together into one track, leaving the vocals out. You can use our Instrumental Maker to upload your file and start the separation.
How do I get instrumental from song files using AI?
To get the instrumental from a song, you must upload an audio file to the servers. The system accepts five specific audio file formats. You can upload an MP3, a WAV, a FLAC, an M4A/AAC, or an OGG file. The file you choose to upload must be stored on your device. The system does not accept links to external sites, so you cannot paste a link from YouTube or a similar video platform to extract the audio. Your audio file must also fit within the size and length limits. The maximum file size you can upload is 90 MB. The maximum length for the song is 10 minutes. If you have a file that meets these requirements, you can upload it to the Vocal Remover or the Karaoke Maker to separate the stems.
What happens during the stem separation process?
The time it takes to process the audio depends on the length of the track and the preview generation. When you upload a file, the servers start processing the audio to create a 30-second preview. This 30-second preview usually comes in about a minute. The preview allows you to hear the results of the separation before you decide to download the files. If you choose to process the entire track, the total time depends on the duration of the original file. For example, a four-minute song is fully split into the five stems in about three minutes. The processing happens entirely on our own servers using the Demucs htdemucs model from Meta AI research.
How does the free preview system work?
Splitting and previewing your audio files are free on the site. Every split you perform plays a 30-second preview of all five stems. This means you can listen to the vocals, the instrumental, the drums, the bass, and the other sounds separately for that 30-second segment. For most songs, this 30-second preview starts exactly at the 0:30 mark of the track. If you upload a song that is under a minute long, the 30-second preview will start from the beginning of the track at 0:00.
There is a limit to how many songs you can split and preview for free each day. If you use the site without an account, you can split and preview 1 song a day. If you create an account and sign in, your limit increases, and you can split and preview 5 songs a day. You can use these free previews on the Acapella Extractor to hear the isolated vocals or the instrumental parts before spending any credits.
What do you get when you download the stems?
When you are satisfied with the 30-second preview, you can choose to download the whole song. Downloading a whole song gives you all five stems. You will receive the isolated vocals, the instrumental, the isolated drums, the isolated bass, and the isolated other sounds. As noted, the instrumental track is the drums, bass, and other added together.
These stems are provided in two different audio formats. You receive the MP3 versions, which are encoded at 320 kbps. You also receive the WAV versions, which are uncompressed 16-bit, 44.1 kHz files. Both formats are included when you download a whole song.
| Specification | Detail |
|---|---|
| Accepted File Types | MP3, WAV, FLAC, M4A/AAC, OGG |
| Maximum File Size | 90 MB |
| Maximum Song Length | 10 minutes |
| Output Stems | Vocals, Instrumental, Drums, Bass, Other |
| MP3 Quality | 320 kbps |
| WAV Quality | 16-bit, 44.1 kHz |
| Download Cost | 125 credits ($0.50) |
How much does it cost to get instrumental from song downloads?
While splitting and previewing are free, downloading a whole song costs 125 credits. The cost of 125 credits is equal to $0.50. You cannot buy exactly 125 credits at a time. Instead, the site offers a one-time purchase of $5. This $5 purchase adds 1,250 credits to your account. Because each download costs 125 credits, the 1,250 credits give you enough for exactly 10 songs.
There is no subscription required to use the site or to buy credits. The $5 payment is a one-time transaction. All payments are processed through FloatPass. The credits you buy are only spent on a download. You do not spend credits to upload a file, to split a file, or to listen to the 30-second preview. You can view more details on the Pricing page.
What happens to my files after I upload them?
Privacy and file management are handled automatically by the servers. When you upload an MP3, WAV, FLAC, M4A/AAC, or OGG file, it is stored temporarily on our own servers to perform the Demucs htdemucs separation. Uploads are deleted after 1 day. The stems generated from your upload are kept slightly longer so you have time to download them. The stems are deleted after 7 days.
During the time your files are on the servers, they are only available to the person who uploaded them. No other users can access your uploaded songs or your separated stems. Users must have the right to process the audio they upload to the site. If there are any copyright notices, they go to contact@stemsplitai.com.
Understanding the instrumental stem and individual tracks
When you get the instrumental from a song on the site, you are receiving a specific audio file. The separation process uses the Demucs htdemucs model from Meta AI research to isolate four core elements: vocals, drums, bass, and other. The "other" category includes instruments like guitars, keyboards, synthesizers, and strings. The instrumental stem is not generated as a separate extraction from the AI model. Instead, the instrumental track is created by adding the drums, the bass, and the other stems together into a single file. This means the instrumental file contains all the music from the original track, with only the vocals removed. You receive this combined instrumental file along with the individual drums, bass, and other files when you spend 125 credits to download the whole song.
When you decide to make a purchase, the transaction goes through FloatPass. The $5 one-time purchase is designed to be straightforward, with no subscription attached to your account. Once the payment clears through FloatPass, the 1,250 credits are added to your balance. You can then spend 125 credits to download the five stems for a single song. If you split a song and listen to the 30-second preview but decide not to download it, you keep your credits. Credits are only spent on a download. This system ensures you only pay for the full MP3 (320 kbps) and WAV (16-bit, 44.1 kHz) files you actually want to keep on your device.
The free tier allows anyone to test the open-source Demucs htdemucs model on their own audio. If you visit the site and do not sign in, you are limited to 1 song a day. This allows you to upload an MP3, WAV, FLAC, M4A/AAC, or OGG file up to 90 MB and 10 minutes long, wait about a minute for the 30-second preview, and listen to the stems. If you need to test more files, signing in increases your limit to 5 songs a day. These daily limits apply only to the free splitting and previewing. You still need credits to download the whole song.
Frequently asked questions
How much does it cost to get instrumental from song files?
Splitting and previewing are free, but downloading a whole song costs 125 credits, which is $0.50. You can make a one-time $5 purchase to add 1,250 credits to your account, which is enough for 10 songs.
What file formats can I upload to get instrumental from song tracks?
You can upload an MP3, WAV, FLAC, M4A/AAC, or OGG song. The file must be up to 90 MB and 10 minutes long, and links such as YouTube are not accepted.
How long does it take to get instrumental from song uploads?
A 30-second preview usually comes in about a minute. If you process a four-minute song, it is fully split in about three minutes.
What audio quality are the downloaded stems?
When you download a whole song, you receive all five stems as MP3 files at 320 kbps and WAV files at 16-bit, 44.1 kHz.
How many songs can I split for free each day?
You can split and preview 1 song a day without an account. If you are signed in, you can split and preview 5 songs a day.
Are my uploaded files kept private?
Yes, files are only available to the person who uploaded them. Uploads are deleted after 1 day and the separated stems are deleted after 7 days.