What is Type Studio?
A music analysis and generation app. Hand it one song and it reads it, maps its structure, or rebuilds the sound. Describe a song in text and it generates original audio to play, share, or iterate on. It exists because most music tools tell you what you played, not how a track is built or how to make one like it.
How do I sign in?
Email and password through Supabase Auth. There are no streaming accounts to connect, hence nothing to authorize. Sign-in is the only credential, so keep it private and sign out from devices you no longer trust.
What are the modes?
Eleven operations. Typer Intelligence generates a grounded analysis of a single song. Typer Ears turns a reference song back into a prompt and remakes it. Typer Explore gives a detailed breakdown of one song. Typer Vision runs an iterative loop to refine a fresh track inspired by a reference. Song Structure maps a song's sections, tempo, and key. Compare Two Songs lines two tracks up and shows what's different. Music is Prediction maps cognitive expectation curves and musical chills triggers. Create describes a song and generates it from scratch. Audio Static Views Generator draws a song's full static analysis as one downloadable view. MIDI Render plays a song or poster as the exact notes the analysis transcribed. Special Explore measures the physics of one song, returning five tracks that match it.
What's the 360 view?
A single static portrait of a whole song: its form and sections, the instruments over time, the full note roll, the chord progression, the frequency map, a self-similarity matrix, the key, the groove, and the timbre fingerprints, all on one page. Audio Static Views Generator renders it, and Typer Vision's loop generates audio that climbs back toward that view, so the 360 doubles as both a read of a song and a target to rebuild it against.
How does the generation work?
Three surfaces, each for a different intent. Create runs Google's Lyria against your text prompt, one prompt in, one 30-second clip out. Typer Ears and Typer Vision run Lyria or ACE-Step inside a Gemini-driven loop, grading each round against a reference with DSP measurements (tempo, harmony via chroma, timbre via MFCC, brightness via spectral centroid) and refining the prompt until the output matches. MIDI Render takes a song's transcribed notes and plays them through a FluidSynth soundfont, which is why its output reproduces the exact melody rather than reinterpreting it.
What models power it?
Google Gemini on Vertex AI handles the analysis, the reasoning, and the optimizer critic. Google's Lyria and Meta-lineage ACE-Step (hosted on Modal) generate the audio. Typer Intelligence's deep-research dossier runs Gemini with Google Search so it can ground its claims against real web sources and cite them.
How accurate is it?
Best-effort. Gemini is strong on the editorial read, weaker on long-tail or very new releases and on crediting features. Generated audio lands a tight match on genre and mood but can drift on tempo, key, and instrumentation, which is why Typer Ears and Typer Vision exist. A single-shot Create clip matches on the first try maybe 60 to 70 percent of the time; a few iterative rounds usually push the composite similarity into the high 80s or low 90s.
Is my data private?
Yes. There are no analytics SDKs in the app, nothing is sold, and nothing trains a model, including your generated audio. Every external call originates from the backend and carries only what the provider needs to do its job. The full breakdown is on the privacy policy page.
Can I delete everything?
Yes. Every run, render, reconstruction, and optimizer job has a delete control. For a full wipe of everything keyed to your account, email m@typeaistudio.com from the address you signed in with.
Does it cost anything?
It's free to start. Beyond a small number of complimentary generations, Type Pro is 9.99 US dollars per month (or the store's local equivalent) through the Apple App Store or Google Play. The store handles the charge, so Type Studio never sees your card.
Can I use a generated track commercially?
Not without your own rights-clearance pass. AI output can resemble existing material, and Typer Ears and Typer Vision, which work from a reference you provide, can closely rebuild that source. A MIDI Render reproduces the exact transcribed notes, so clear the underlying composition before distributing one built from someone else's song. The full position is in the terms of use.
More questions: m@typeaistudio.com. See the privacy policy, the terms of use, and support.