Formats and Limits
What Aimusicis accepts as an upload, how long and how large a file can be, which model versions the generator offers, and what each separation returns.
The figures on this page come from the same code that enforces them in the upload picker, the generator form and the credit table, and a test pins each one to its source — so what is written here is what the tool actually does.
Music generator and separation uploads
| Limit | Value |
|---|---|
| Accepted formats | MP3, WAV, OGG, M4A, FLAC and AAC |
| Maximum file size | 50 MB per file |
| Separation file size | 90 MB per file |
| Maximum length | 15 minutes of audio |
Anything the picker does not list has to be converted before upload. This table describes the vocal-remover, stem-splitter and generator reference-track picker: the generator's reference track stops at 50 MB, while the vocal remover and stem splitter take files up to 90 MB, which fits about nine minutes of 16-bit stereo WAV. Specialized workbenches state their own limits.
Generator model versions
The AI Music Generator supports five model versions: Music V2.5, Music V2, Music V1, Music Solo and Music Studio. The version you pick sets how many songs one run returns, how long each text may be and which controls the form shows; a generation with or without vocals costs the same at a given version (see Credits and Plans).
| Version | Songs per run | Description (prompt mode) | Lyrics (custom mode) | Style (custom mode) | Length |
|---|---|---|---|---|---|
| Music V2.5 | 2 | up to 500 | up to 5,000 | up to 1,000 | set by the song |
| Music V2 | 2 | up to 500 | up to 5,000 | up to 1,000 | set by the song |
| Music V1 | 2 | up to 500 | up to 5,000 | up to 1,000 | set by the song |
| Music Solo | 1 | 10 to 2,000 | up to 3,500 | 10 to 2,000 | set by the song |
| Music Studio | 1 | up to 4,100 | up to 4,100, at most 30 sections of 30 lines | up to 1,000 | chosen: 3 seconds to 600 seconds |
Text limits are in characters. Only Music V1, V2 and V2.5 take a persona, a vocal gender and the style and weirdness sliders, and only they can extend, cover, mash up or edit a song; a song made with Music Solo or Music Studio plays, downloads and shares like any other.
Lyrics generator inputs
The AI Lyrics Generator takes six inputs: a title, a prompt, and a choice of mood, structure, style and language. It draws 2 credits per run.
What a separation returns
A vocal split writes two files from one mixed track — the isolated vocal and the instrumental behind it — with tempo and key untouched, so both halves drop back onto a timeline together. The stem splitter writes individual stems instead, with drums and bass apart from the rest. Neither run changes the length of the audio.
AI Mastering
| Limit | Value |
|---|---|
| Accepted formats | MP3, WAV, OGG, M4A, FLAC and AAC |
| Maximum file size | 180 MB per file |
| Maximum length | 15 minutes of audio |
AI Mastering takes a target mix and a reference track, then matches your track's tone and loudness to the reference. It keeps the stereo image and returns a mastered WAV with an A/B compare against the original. The form shows the credit cost before you run it.
Browser MIDI, audio editing and analysis
| Browser tool | Input limit | Export |
|---|---|---|
| MIDI Editor | 5 MB MIDI files | MIDI, WAV or MP3 |
| Pitch and slowed/reverb editor | 90 MB and 15 minutes of audio | WAV |
| Audio to MIDI | 90 MB and 15 minutes of audio | MIDI |
| Key and BPM analysis | 20 MB and 10 minutes of audio | CSV |
The MIDI Editor edits notes, tracks, velocity and tempo. It can load a custom SoundFont, and keeps a browser autosave on this device. Export a MIDI copy before moving your project to another device.
Audio Pitch Changer and Slowed and Reverb Generator support file input, microphone recording and sample tracks. They change speed and pitch independently and add reverb. Browser codec support determines which audio files can be decoded.
These browser tools draw zero generation credits.
Audio to MIDI transcribes pitched recordings into editable notes and can open them in the MIDI editor. Key & BPM Finder, Beat Key Finder and Beat Finder estimate key, Camelot notation and tempo, and read energy and danceability from the same analysis. They accept multiple files for sequential analysis and export a CSV table. These browser tools process audio on your device without credits; their results are not saved to My Music. Musical estimates need checking by ear, especially on dense, atonal or changing-tempo material.
Lofi Converter renders a 0.94×, low-pass-filtered version with vinyl-style noise. It accepts up to 90 MB and 15 minutes of audio and exports stereo 16-bit WAV at 22,100 Hz. Sign in to save the source and result in Conversion history. Public example previews run without uploading or saving files.
Chord Maker and Chord Progression Generator share twelve keys, major/minor mode, ten styles and piano/guitar previews. Each chord takes one beat at 120 BPM in the MIDI export. Open the sequence in the MIDI editor to change notes, voicing, length or tempo. These local progressions do not consume generation credits or create My Music history entries.
Text to MIDI composition
AI MIDI Generator creates an editable score from a text prompt. Choose genre, mood or instrument tags, then set BPM, key, time signature and bar count. Sign in to generate and find saved scores in its Results tab. Preview or export MIDI there, or open the score in the browser MIDI editor. Editor changes stay on this device; export them to keep a revised file. The form shows the credit cost before generation. Original sample scores can be previewed, exported and edited without generating a new result.
What is not available yet
Tools behind a "coming soon" notice are not available yet. They draw no credits because they cannot run. No tool page carries that notice today: the music-video makers and the MIDI tools above all run.
Next: Beginner Guide to Generating Songs or Credits and Plans.
For a new song, choose from Music V2.5, Music V2 and Music V1. Music V2.5 is the starting choice when you know your direction; Music V2 helps explore alternatives; Music V1 helps sketch ideas. Projects using a retired model must choose a current model before generating again. Existing songs remain available in the library. Advanced editing tools show only the models their operation accepts.