Why most answers say you need two sites
AI song generators and stem splitters grew up as separate products. Song generators such as Suno and Udio focus on writing and singing the track; stem splitters such as LALAL.AI, Moises or Ultimate Vocal Remover focus on taking finished audio apart. On many song generators, stems or even plain downloads sit behind a paid plan, so the usual advice is to generate on one site, download, and upload the file to a second tool.
That works, but it costs time and often needs two accounts. TuneSplit was built as one studio: the song generator, the stem splitter, the vocal remover, the BPM and key finder and the lyric video maker share one page layout and one free credit balance, and none of them needs a sign-up.
What the free tier covers
Every UTC day each browser gets 30 free credits and 3 free stem-splitter files. Credits reset at 00:00 UTC. Here is what that buys for the song-plus-stems workflow:
| Step | Free tier | Cost in credits | Output |
|---|---|---|---|
| Generate a song | Up to 3 generations a day | 10 per generation (2 versions) | MP3, up to 2 minutes, personal use |
| Stems preview of your song | First 30 seconds, 1 per song version | 0 | 4 stems: vocals, drums, bass, other |
| Full stems of your song | Yes, from your daily credits | 20 | 4 full-length WAV stems |
| Split an uploaded file | 3 files a day, up to 10 minutes each | 0 (then 30 per started 5 minutes) | 2 or 4 stems as MP3 |
So one free day can be 1 generation (10 credits) plus full WAV stems of one version (20 credits), which uses exactly 30 credits. If you would rather keep credits for more songs, download the MP3 of a version and run it through the stem splitter, which uses one of your 3 free files instead of credits.
Step by step, with the real buttons
1. Write the prompt. A good prompt names the mood, genre, tempo and subject, for example: upbeat indie pop about a road trip with friends, female vocal, around 110 BPM. If you already have lyrics, paste them and the song is sung from your words. The lyrics generator can write a first draft for you, with five free lyric generations a day.
2. Compare the two versions. Every generation runs two different music models, so version A and version B usually differ in melody, arrangement and voice. Play both before you pick one to split.
3. Press Stems. TuneSplit runs Demucs (htdemucs), the open-source separation model from Meta research, on the first 30 seconds of that version. You get four players: vocals, drums, bass and other. Use the preview to check the split is clean before spending credits.
4. Get the full stems. Press Get full-length WAV stems. 20 credits are taken from your balance, and if the job fails they go back automatically. A three-minute song usually takes 20 to 60 seconds.
5. Download. Each stem has its own download button. WAV files are lossless, so they are the better choice for mixing or remixing.
What people do with the stems
- Make a karaoke or backing track of their own AI song by muting the vocals stem.
- Keep only the vocals and put them over a beat they made in a DAW.
- Replace the AI drums with their own drum loop while keeping the AI vocals and bass.
- Re-pitch or re-time a stem with the pitch changer or tempo changer to fit another project.
- Make a lyric video from the finished song with the free lyric video maker on the same site.
Limits to know before you start
- Free songs are up to 2 minutes and are for personal, non-commercial use; a paid plan adds longer songs and a commercial licence.
- Stems come as 4 parts: vocals, drums, bass and other. There is no separate guitar or piano stem.
- Separation is an AI estimate, so a little bleed between stems can remain on dense mixes.
- Uploaded files for the stem splitter are limited to 10 minutes each on the free tier and are deleted within 24 hours.
- Credits and free-file counters are kept per browser, not per account, so clearing site data resets your history.
No voice cloning and no AI covers of real singers are offered. Only upload audio you have the right to use.
Frequently asked questions
Is there one free website that makes an AI song and splits it into stems?
Yes. TuneSplit generates 2 versions of a song with vocals and then splits a version into vocals, drums, bass and other on the same page. A 30-second stems preview is free, full WAV stems cost 20 of the 30 free daily credits, and no sign-up is needed.
Can I download the AI song on the free tier?
Yes. Each finished version has a download button that saves an MP3 of up to 2 minutes for personal use.
Do I need to upload my AI song to another tool to get stems?
No. Press Stems under the version you want. If you prefer, you can also download the MP3 and drop it into the TuneSplit stem splitter, which is free for a few files a day.
What format are the stems?
Full stems of a song you generated are WAV. Stems from the uploaded-file splitter are MP3.
Which models does it use?
Songs come from MiniMax Music v2 and ACE-Step, one version each. Stems come from Demucs (htdemucs).