I've been building MusiCraft ā an AI music platform that puts stem splitting, music generation, mastering, and a bunch of other audio tools into one browser app. No installs, no DAW plugins, everything runs online.
I want to walk through what the platform actually does, because "AI music tool" can mean anything from a toy demo to a full production suite. This is closer to the second one.
The core idea
Most AI music tools do one thing. There's a stem splitter over here, a text-to-music generator over there, a mastering service somewhere else. You end up with six tabs open, uploading the same file to three different sites, downloading and re-uploading between them.
MusiCraft puts all of these in one workspace. You sign in once, your files live in one place, and the tools talk to each other. Split a track into stems, then master the vocal stem, then generate a new instrumental to go under it ā all without leaving the app.
What's inside
Generation tools
Text-to-Music is the main one. You describe what you want ā genre, mood, instruments, tempo, lyrics or no lyrics ā and the AI produces a full track. Each generation returns two variations so you can pick the one that fits better or use both.
But one generic generator doesn't cover everything. Lo-fi beats need a completely different feel than a rock track or a K-pop title track. So I built dedicated generators for specific genres: Lo-fi, R&B, Rock, Hip-Hop, K-pop, Jazz, Phonk, and more. Each one is tuned to the conventions of that genre ā the right drum patterns, the right harmonic choices, the right production style.
There's also:
- Jingle Maker ā short-form music for ads, intros, podcasts. You describe the brand, the mood, the length, and it writes a jingle.
- Hum to Song ā record yourself humming or singing a melody, and the AI turns it into a fully produced track.
- Photo to Music ā upload an image and it analyzes the color palette, lighting, and composition to generate a mood-matched soundtrack. Sounds gimmicky, but content creators actually use this a lot for social videos.
- Text to Speech ā turn a script into natural spoken audio.
You can try the full music generation suite on the AI Music Producer page.
Processing tools
Stem Splitter ā this is probably the most-used tool on the platform. Upload any song and it separates it into individual stems: vocals, drums, bass, guitar, piano, etc.
What makes it different from the usual Demucs wrapper: it has 20+ separation modes. Most stem splitters give you one option ā 4-stem (vocals, drums, bass, other). That "other" bucket is where your guitar, piano, synths, and everything else gets mashed together.
MusiCraft lets you choose exactly what you need:
- Vocal/Instrumental ā cleanest vocal isolation, two outputs
- 4-stem ā the classic split
- 6-stem ā adds guitar and piano as separate tracks
- 8-stem ā full band, lead vocals separate from backing vocals
- Drum Kit ā just the percussion, broken into kick, snare, hi-hat, cymbals
- Single instrument modes ā isolate just the piano, just the guitar, just the sax, just the strings
- Choir mode (SATB) ā separate soprano, alto, tenor, bass vocal parts
- Guitar lead/rhythm ā split a guitar track into lead and rhythm parts
Each mode has its own landing page that explains what it isolates, what the trade-offs are, and lets you upload directly from the page.
AI Mastering ā upload a mix and get a mastered version back. You can provide a reference track and the AI matches the loudness, EQ curve, and dynamics to it.
Denoise ā remove background noise, hiss, hum from recordings.
Dereverb ā reduce room echo and reverb for drier audio.
Enhance ā restore quality lost to MP3 compression or low-bitrate encoding.
Analysis tools
BPM & Key Finder ā detects tempo and musical key from any track, right in the browser.
Audio to MIDI ā converts audio (piano, vocals, drums, general melodies) into MIDI files you can import into your DAW.
Music Copyright Checker ā identify a song from a clip or check if your track has potential copyright issues before you publish.
The agent
Beyond the individual tools, there's an AI agent that ties everything together through a chat interface. You can talk to it in natural language:
"Take this track, split the stems, remove the vocals, master the instrumental, and generate an album cover that matches the mood."
It chains the tools together and runs the whole workflow. Think of it as a producer assistant that knows how to use every tool in the platform.
Who actually uses this
When I started building MusiCraft, I assumed the audience would be bedroom producers. Turns out the user base is way more diverse:
- Content creators who need background music for YouTube, TikTok, podcasts ā they use text-to-music and photo-to-music constantly
- DJs and remixers who need clean stems for mashups and live sets
- Music teachers and choir directors who use the SATB choir mode to create practice tracks for individual voice parts
- Podcasters who need jingles, intros, and cleaned-up audio
- Cover artists who want karaoke-style instrumentals from any song
- Game devs and filmmakers who need royalty-free soundtracks that match specific moods
The commercial licensing is straightforward ā tracks generated on paid plans are royalty-free for streaming, sync, monetization, whatever. No per-track fees, no revenue sharing.
Pricing
The platform runs on a credit system. Every tool costs credits to run, and plans come with monthly credit pools:
- Free ā 4 songs/month to try it out
- Basic ā 80 songs/month
- Pro ā 200 songs/month
- Studio ā 600 songs/month
- Business ā 2,000 songs/month
Annual plans save 20ā25%. Stem splitting, mastering, and all the processing tools use credits from the same pool ā one subscription covers everything.
What's next
I'm still actively building. Current priorities:
- More separation modes for less common instruments
- Better vocal cloning (train a voice model from 30 seconds of audio)
- Workflow templates ā pre-built chains like "remix this song in a different genre"
- More language support ā the platform already runs in 9 languages (EN, ES, PT, JA, KO, DE, ZH, TW, FR) but I want to expand
Try it
Free account with starter credits, no card required: musicraft.ai
Happy to answer any questions about the platform. If you work with audio in any capacity, I'd especially love to hear what tools or workflows you wish existed.
