Transcribing a song thoroughly is hard and time-consuming. If you are only interested in the main melody and chords, the task becomes easier, but what if you need every part of the track for a perfect reconstruction?
Let's face it: most of the transcriptions out there are wrong. Take any track, look it up on a tab site and you'll find a variety of transcriptions. Some are very simple: chords only with no indication of rhythm, some are full reconstructions done with a software like Guitar Pro. The quality can be anything, and if you have any musical knowledge you'll notice who knows what they are doing and who is just guessing. (/rant)
If you're analyzing a popular song, you'll have a better chance finding a decent transcription. It's still a good idea to verify it yourself.
Here are the things that will help you analyze and listen to your track for that perfect transcription or recreation.
1. Use a good pair of speakers or headphones
Listening requires good equipment. Use high quality speakers, monitors or headphones that have a good resolution, meaning that you'll be able to distinguish between different parts in the mix better. Don't forget the rest of the chain either: you'll need a good amp or audio device also. If you're into music, you should have these things sorted out already.
2. Use your DAW/audio editor
Import the track in your DAW or use an audio editor (e.g. Audacity) where you can easily skip around, start and stop repeatedly, loop parts and apply effects.
You can analyze a song from your music player also but you'll get frustrated pretty fast if there are difficult parts that you need to listen over and over again.
3. Use a similar sound when transcribing a part
It makes sense to use the same instrument when you are figuring out a part in the song. If there's a piano, use a piano sound to play along. You could also try to tweak your sound to match the original with EQ and effects.
If you're working in a DAW, record or manually program your best guess of the part and start A/B comparing it with the original via muting and soloing. If your sound is similar to the original, you can hear the differences easily and iterate your guess.
If your goal is to recreate the whole track, transcribing and recreating can be done in this same step. You just need to figure out the tempo of the original to perfectly align with your DAW grid.
4. Use EQ/filters
If you're analyzing the bass, apply a low-pass filter. If you're analyzing the hihat, apply a high-pass filter. If you're analyzing the vocals, apply a band-pass filter or an EQ boost in the mid frequencies. Focus on the stuff you are listening to and try to filter out the other things.
Sometimes you'll need very sharp filters. One example is to isolate the bass drum from the snare drum. They can share a lot of frequencies so it might be difficult to tell if there is a bass drum under a certain snare hit or not. In this case, take a very sharp (high Q or resonance) filter and focus on the boom of the bass drum.
 |
| In iZotope's Ozone you can sweep frequencies with a narrow band easily by pressing down Alt while using mouse. |
5. Use mid/side channels
Using different channels might work if you have hard panned elements. Listen to left and right separately. One extra trick is to use mid/side processing.
Middle and side channels are just a different way of encoding the usual two channels: left and right. The mid channel is the sum (L + R) and the side channel is the difference (L - R). If you think about this for a while, you'll notice that the mid channel amplifies sounds that are common between left and right (i.e. centered things) and the side channel amplifies sounds that are different (hard panned or stereo-effected things).
Fun fact: simple vocal removers (that usually don't work) use mid/side processing by reducing the middle channel. This works on the assumption that the vocals are centered, but is also reduces any other centered material in the mix.
Some parts that are difficult to analyze in the original mix might be very audible in the side channel.
 |
| In Ozone you can solo the mid/side channels separately |
6. Use time stretching
If you have fast or difficult parts, slowing the whole thing down might make it easier to analyze. Be sure to use time stretching (which retains pitch but changes speed), not resampling (which alters both).
For a good result, you may need to tweak the time stretching parameters or the algorithm used. For example Ableton Live has different algorithms for drums (better transients), tones and complex material.
7. Use a spectrogram
Instead of using your ears it's sometimes better to use your eyes. If you struggle with identifying note pitches or the number of rhythmic hits, use a visual analyzer.
 |
| In a spectrogram (like Sonogram SG-1 here) you may see things like drum hits and melody lines more easily. |
A spectrogram visualizes the frequency changes over time, so you might see things like melodies and drum rhythms more easily. Analyzing the picture requires a keen eye though: because most of the sounds have a timbre that spans a lot of ground frequency-wise (i.e. they are not composed of a single frequency component like a sine wave), you won't see a single line but many repeated lines.
A spectrogram may be useful where you need to compare relative pitches a few hits or bars apart. If they occupy the same horizontal line in the graph, they are the same note.
8. Apply your knowledge of music theory and different instruments
If you know at least something about music theory, analyzing songs becomes so much easier. Knowing basic things like keys (circle of fifths), chords, scales, common chord progressions (I - IV - V) etc. will help you immensely. Know the difference between a G sharp and an A flat.
With your musical knowledge you'll be able to "fill in the blanks" without analyzing every single note. You'll recognize some chords and melodies immediately. Transcribing becomes faster.
If you know how to play a specific instrument, you'll be better at analyzing it. Learn about different instruments: how they are usually played, what kind of sound do they produce, what is their range etc. Learning to play more than one instrument is never a bad idea.
Transcribing a song is a skill that improves by doing and practicing (no surprise). If you are a beginner, learn some theory basics and start with a simple song. Study the transcriptions made by others (remember that most of them are wrong, find the errors). Listen, sing along, play.
