No two instruments transcribe the same way. A piano stacks many notes at once, a bass sits low where it is easy to miss, a saxophone slides between pitches, and drums carry rhythm with no pitch at all. The right approach depends on what you are listening to, and the most reliable results come from transcribing one instrument or part at a time. This guide covers how to transcribe each instrument, from piano and guitar to horns, strings, voice, and drums, and where AI does the heavy lifting so you are editing a real score instead of starting from a blank page.
Find your instrument
Pick the instrument you are transcribing, read its guide, and take the next step.
| Instrument | Guide to read | Next step |
|---|---|---|
| Piano (any style) | Transcribe piano music with AI | Try Songscription free |
| Classical piano | Transcribe classical piano | Upload a recording |
| Gospel piano | Transcribe gospel piano | Try Songscription free |
| Guitar solo | Transcribe a guitar solo | Upload a recording |
| Fingerstyle guitar | Transcribe fingerstyle guitar | Try Songscription free |
| Bass line | Transcribe a bass line | Upload a recording |
| Saxophone melody | Saxophone melody to sheet music | Try Songscription free |
| Trumpet melody | Trumpet melody to sheet music | Upload a recording |
| Vocal melody | Vocal melody to sheet music | Try Songscription free |
| Violin | Audio to sheet music for violin | Upload a recording |
| Jazz solo | Transcribe a jazz solo | Create a free account |
| Drums | Transcribe drums with AI | Create a free account |
Why one instrument at a time
The single most useful habit in transcription is to work one instrument at a time. Focusing the model on one part gives a faithful, editable transcription of that line rather than a rough guess at everything in the mix. Songscription is built around this: you transcribe a single instrument or part, and the models are generally robust enough to isolate the instrument you want directly from a full mix, so you do not need to split the song into stems first. On an especially dense or crowded track, optional stem separation can make a part cleaner still. To score a whole arrangement, you transcribe each part and combine them, which is covered in the full-band guides. The payoff is accuracy: each part comes out clean enough to read, edit, and hand to a player.
Piano, guitar, and bass
Polyphonic instruments, where several notes sound together, are where AI transcription is strongest. Piano is the clearest case: transcribing piano music with AI walks the general workflow, while classical piano and gospel piano cover the demands of those styles. On guitar, a single-line lead and a fingerpicked part are different jobs, handled in transcribing a guitar solo and transcribing fingerstyle guitar, with the broader instrument-specific picture in our guitar transcription guide. And because the bass lives at the bottom of the mix where it is easy to lose, transcribing a bass line covers how to pull it out cleanly.
Horns, strings, and voice
Single-line melodic instruments are mostly about capturing one expressive line accurately, including the slides, bends, and phrasing that make it sound human. The workflow for a horn is in turning a saxophone melody into sheet music and turning a trumpet melody into sheet music, with the transposing-instrument wrinkle that those parts get written in their own reading key. For strings, audio to sheet music for violin covers what works, and a sung line is its own case in turning a vocal melody into sheet music. A jazz solo on any of these is the deep end, covered in transcribing a jazz solo.
Drums
Drums are the outlier: there is no pitch to read, so the notation is about which drum is hit and when, written on a percussion staff rather than a melodic one. That changes what to expect from a transcription and how to read it, which is exactly what transcribing drums with AI covers, including where the result is solid and where you will still tidy up by hand.
Transcribe the part you actually need
Upload a recording, point Songscription at the instrument you care about, and get editable notation back in minutes. The free tier is enough to transcribe your first part.
Related guides
- Keys and frets: piano, classical piano, gospel piano, guitar solo, fingerstyle guitar, and bass line.
- Horns, strings, voice: saxophone, trumpet, violin, voice, and jazz solo.
- Drums: transcribe drums with AI.
Frequently Asked Questions
Can AI transcribe any instrument?
AI transcribes most pitched instruments well, and it handles polyphonic instruments like piano and guitar especially well, where many notes sound at once. Each instrument is its own problem: a bass is low and easy to mask, a saxophone bends and slides between pitches, and drums have no pitch to read at all. Songscription transcribes one instrument or part at a time from a recording, so you point it at the instrument you care about and get editable notation back rather than a guess at the whole mix at once.
How do I transcribe one instrument out of a full song?
Work one instrument at a time. Songscription models are generally robust enough to isolate the instrument you want directly from a full mix, so you point the tool at one part and it pulls that line out of the recording on its own, without splitting the song into stems first. On an especially dense or crowded recording, optional stem separation can make a part cleaner still, but it is an aid rather than a requirement. Songscription is built around transcribing a single instrument or line to editable notation, so the reliable workflow is to transcribe the part you want, then repeat for any other parts you need and combine them.
What instrument is hardest to transcribe?
Drums and heavily processed or pitch-bending parts are the hardest. Drums carry rhythm and timbre but no pitch, so they are notated differently from melodic instruments, and a saxophone or a guitar with heavy effects can blur the exact pitch boundaries the model relies on. These still transcribe usefully; you just review and clean up more than you would for a clean solo piano recording, which is why an editable result matters.
Does Songscription transcribe multiple instruments at once?
Songscription transcribes one instrument or part at a time rather than splitting a full band into separate parts in a single pass. That is a deliberate trade: a per-instrument pass gives each part a clean, accurate, editable transcription instead of a rougher all-in-one result. The models are generally robust enough to isolate the instrument you want directly from a full mix, so you do not need to split the song into stems first. To score a full arrangement, you transcribe each part and assemble them.
The fastest way to start is on a part you already want to read. Upload a recording with Songscription and transcribe your instrument into editable notation.
