Audio (MP3/WAV) to Tablature Converter

[RedHotFuzz]

I asked a similar question recently and got many of the expected "HOW DARE YOU NOT TRANSCRIBE BY HAND LIKE THE ARTISANS OF OLD!" responses (both here and on Reddit - the Reddit thread got so obnoxious I finally just deleted the post). I'm still on the hunt. I discovered Logic Pro's stem splitting feature and that's been great for isolating bass tracks. Results vary but are usually pretty good. I've tried a few AI transcription services and those are mostly a giant fail, but Songsterr isn't too bad. I didn't realize I could try the stems > MIDI > notation approach in Logic Pro as mentioned above. I'll have to give that a shot.

You'd think this would be a fairly simple task for AI, but it's not there yet. It often gets the notes right, but has no concept of how to logically select fret positions that would make sense for efficient playing.

 

[TightGroove70]

I'm looking for the same thing that the bass covers on YouTube feature with the tabs moving across the screen during the videos.

Not sure if this was answered. I did not read the whole thread. The YouTubers transcribe those scrolling TAB videos with an app called GuitarPro 8 (not sure if that's the latest version). Then they use video editing software to have it scroll to the tempo of the tune. Since they transcribe it, they can then go ahead and offer the TAB files separately for purchase.

That being said, some of these transcriptions might not be 100% in reference to hand placement/choice of string selection. I find myself using the notation and having to remember that I've chose to play a part differently than what is notated. IMO that works when you initially decide to make a change, BUT having to remember for the next time u play that song, you basically have to "relearn" the part that you decided to alter.
As for right now, AI is not a valid shortcut, and you know what they say about short cuts. I bit the bullet and learned how to use GuitarPro.
It's also great if you want to notate original exercises, licks, sections that you can pull up and access later.

 

[markbmbass]

Not sure if this was answered. I did not read the whole thread. The YouTubers transcribe those scrolling TAB videos with an app called GuitarPro 8 (not sure if that's the latest version). Then they use video editing software to have it scroll to the tempo of the tune. Since they transcribe it, they can then go ahead and offer the TAB files separately for purchase.

That being said, some of these transcriptions might not be 100% in reference to hand placement/choice of string selection. I find myself using the notation and having to remember that I've chose to play a part differently than what is notated. IMO that works when you initially decide to make a change, BUT having to remember for the next time u play that song, you basically have to "relearn" the part that you decided to alter.
As for right now, AI is not a valid shortcut, and you know what they say about short cuts. I bit the bullet and learned how to use GuitarPro.
It's also great if you want to notate original exercises, licks, sections that you can pull up and access later.

Thank you. Does GP8 actually creat the tab from an audio file or do you have to do that yourself?

 

[TightGroove70]

Thank you. Does GP8 actually creat the tab from an audio file or do you have to do that yourself?

I create them myself. Not sure if the functionality exists to import audio.

If you try to transribe them yourself, IMO, there's a learning curve and have to learn the keystroke shortcuts to speed up the process.

 

[markbmbass]

I create them myself. Not sure if the functionality exists to import audio.

If you try to transribe them yourself, IMO, there's a learning curve and have to learn the keystroke shortcuts to speed up the process.

Ok thank you.

 

[TightGroove70]

I just did a Google search and came up with this:

Workarounds for Automated Tablature. If you need automated transcriptions, you'll have to use external tools first. Many users convert their audio files (.mp3 or .wav) to MIDI files using external audio-to-MIDI converters. You can then import that MIDI file directly into Guitar Pro 8 to generate tablature or standard notation, though minor manual editing will usually be required.

I believe this is similar concept to the YT vid @gtirard posted. GP8, is not free, so if you find something that is free and can do the same thing, you might wanna start there. GP8 is a robust engraving app so there might be some stuff that free apps won't offer.

 

[RedHotFuzz]

I'm anxiously awaiting this new site. I hope it lives up to its billing.

Elite Bass Tabs – AI Bass Tab Generator from Your Audio Files

Turn your audio into clean, accurate bass tablature using advanced AI analysis. Upload audio files you own or have permission to use. Perfect for bass practice and learning.

elitebasstabs.com

 

[markbmbass]

I'm anxiously awaiting this new site. I hope it lives up to its billing.

Elite Bass Tabs – AI Bass Tab Generator from Your Audio Files

Turn your audio into clean, accurate bass tablature using advanced AI analysis. Upload audio files you own or have permission to use. Perfect for bass practice and learning.

elitebasstabs.com

Likewise, I hope so too. If so, they will have a great business, that's for sure.

 

[gtirard]

Likewise, I hope so too. If so, they will have a great business, that's for sure.

I think the market for bassists is small, and even smaller for bassists who want to learn this way.
But I understand why it's specialised on bass like this, since we're almost the only monophonic instrument.
For once we're lucky to enjoy new tools before the guitarists

 

[gtirard]

I'm anxiously awaiting this new site. I hope it lives up to its billing.

Elite Bass Tabs – AI Bass Tab Generator from Your Audio Files

Turn your audio into clean, accurate bass tablature using advanced AI analysis. Upload audio files you own or have permission to use. Perfect for bass practice and learning.

elitebasstabs.com

It certainly uses the same technology as what I'm doing in Apple Logic Pro (stem separation, tempo recognition, and audio to midi conversion), but in a more streamlined and focussed user interface, and that's great.

But, as I've used the method quite a bit since I've discovered it, it works well so it's a real time saver, but I still have to clean up the mess each time, because it adds weird notes (depending on how clean is the source track), it doesn't take into account ghost notes, bends or glissendos, etc.

elitebasstabs.com is going to be a subscription service, probably with a limited free access and a "pro" account with monthly fees.

1. How much would you be ready to pay each month for the service, if it gets 95% of the job done correctly in average, but you still need to cleanup 5%?
2. How do you expect it to handle the tabs, considering that there's always several possible fingerings for a given track, the choice being a matter of personal preference. Would you expect it to propose only one tab notation corresponding to a "recommended" fingering, or would you expect it to propose several options?

I wish them good luck with their "Early Access on the Horizon" status, as I know by experience that it often takes more time to fix the last 5% remaining bugs than to code the first 95% of an app.

I'm very curious to see how well it works...

 

[markbmbass]

I think the market for bassists is small, and even smaller for bassists who want to learn this way.
But I understand why it's specialised on bass like this, since we're almost the only monophonic instrument.
For once we're lucky to enjoy new tools before the guitarists

The guy seems to have a great idea. Might not make tens of millions but I'm sure he'll find enough of a market with this. Either way, I'm looking forward to it.

 

[RedHotFuzz]

It certainly uses the same technology as what I'm doing in Apple Logic Pro (stem separation, tempo recognition, and audio to midi conversion)

I've been using Logic Pro for stem separation to create play-along tracks, but I'm not familiar with using it to convert audio to actual notation using MIDI. Is this something you've had to learn yourself or are there tutorials out there on this exact process?

 

[gtirard]

I've been using Logic Pro for stem separation to create play-along tracks, but I'm not familiar with using it to convert audio to actual notation using MIDI. Is this something you've had to learn yourself or are there tutorials out there on this exact process?

I think I've found the method by myself, with the help of several Youtube channels.

Basically, once you've done the stem separation and tempo detection, you can use "Flex Pitch" to make an audio-to midi-conversion.

I've made a demo video in this other thread:

Post in thread 'Automatic conversion from MP3 to sheet music: praise for Logic Audio'

Mar 13, 2026

Just a quick & dirty demo of how to go from a full band mix mp3 file to a bass score with tabs:

1. Import the song
2. Find the tempo
3. Stem Spliter
4. Show the audio bass isolated track
5. Show Flex
6. Select Flex Pitch
7. Edit -> Create Midi track from Flex Pitch data
8. Cleanup the track in the piano roll (*)
9. Show the score
10. Add bass tabs

Everything was recorded in real time, it only took me 3 minutes.
(*) In real life, step #8 may require much more time than what I did here

Edit: Let's be honnest, I choose "The Look" from Metronomy because it works...

• gtirard

• 

Since I've made this video, I've used this method for my own needs quite a few times (in order to learn walking bass), it works well.

 

[HotelX]

I use gtirard's method pretty often and it works well.

Also my bandmates love being able to hear their isolated parts. Really allows you to get into the subtle nuances and hear what is actually being played

 

[elitebasstabs]

Just wanted to add a few things to this discussion....I hope I am not acting out of line being a developer, but I would like people to be more informed. There is no software based transcription that is 100% accurate, that includes mine. 90% accuracy seems to be the accepted rate. This is pitch detection accuracy, getting the right note.. But why is bass so hard to get right?

By nature, bass notes are hard to detect. A low E is 41Hz it only vibrates ~41 times a second — a vocal or guitar note might be 200-800Hz — so the software has way fewer clean wave cycles to grab a pitch from, especially on fast lines. There's also a well-known octave problem: a lot of bass tone has a fundamental quieter than its own overtones, so the software locks onto the wrong octave and reports a note that's confidently wrong, not just slightly off. On top of that, bass sits buried under kick drum and low guitar even after isolation, and long sustained notes actually get less clear mid-decay, not just at the attack. Ghost notes, fast repeated plucks, and slides are their own separate headache. Bass sits in the exact frequency range and mix position where every assumption pitch-detection makes gets stress-tested — guitar and vocals have it easy by comparison.

Fingering is its own mess, and it's not really fixable either. Most notes can be played in 2-3 spots on the neck with no single "correct" answer — it depends on speed, what's around it, and personal style. Ask ten bassists to play the same line and you'll get several different, all-correct fingerings back. Software has to pick one and show it to you. Worse, a pattern that's ergonomic for a slow groove is often wrong for a fast passage (using an open string as a quick accent, for example) — so there's no one profile that pleases everyone, because there's no one way real people actually play. I am currently working on these problems, especially the fingering. But even with my 45+ years of experience playing bass I have yet to create a fingering profile that gets it all right. Ive thought about asking all experience levels of players to describe how they come to play the songs they know a certain way, pattern or position and what factors influence these decisions.. its a hard one and again I dont know if I would be stepping out line asking that question here.

GP

 

[PNW Tony]

Just wanted to add a few things to this discussion....I hope I am not acting out of line being a developer, but I would like people to be more informed. There is no software based transcription that is 100% accurate, that includes mine. 90% accuracy seems to be the accepted rate. This is pitch detection accuracy, getting the right note.. But why is bass so hard to get right?

By nature, bass notes are hard to detect. A low E is 41Hz it only vibrates ~41 times a second — a vocal or guitar note might be 200-800Hz — so the software has way fewer clean wave cycles to grab a pitch from, especially on fast lines. There's also a well-known octave problem: a lot of bass tone has a fundamental quieter than its own overtones, so the software locks onto the wrong octave and reports a note that's confidently wrong, not just slightly off. On top of that, bass sits buried under kick drum and low guitar even after isolation, and long sustained notes actually get less clear mid-decay, not just at the attack. Ghost notes, fast repeated plucks, and slides are their own separate headache. Bass sits in the exact frequency range and mix position where every assumption pitch-detection makes gets stress-tested — guitar and vocals have it easy by comparison.

Fingering is its own mess, and it's not really fixable either. Most notes can be played in 2-3 spots on the neck with no single "correct" answer — it depends on speed, what's around it, and personal style. Ask ten bassists to play the same line and you'll get several different, all-correct fingerings back. Software has to pick one and show it to you. Worse, a pattern that's ergonomic for a slow groove is often wrong for a fast passage (using an open string as a quick accent, for example) — so there's no one profile that pleases everyone, because there's no one way real people actually play. I am currently working on these problems, especially the fingering. But even with my 45+ years of experience playing bass I have yet to create a fingering profile that gets it all right. Ive thought about asking all experience levels of players to describe how they come to play the songs they know a certain way, pattern or position and what factors influence these decisions.. its a hard one and again I dont know if I would be stepping out line asking that question here.

GP

Thanks for the clear, informed explanation. I knew some of these things and hadn't considered others. Much appreciated.

 

[elitebasstabs]

You are very welcome, glad it helped

Thanks for the clear, informed explanation. I knew some of these things and hadn't considered others. Much appreciated.

 

[RedHotFuzz]

I've been testing Elite Bass Tabs and have every hope it turns out well because I really, really want an AI bass tab generator that works, but it's surprising how much better the human ear is over AI still. With isolated bass tracks it still gets a lot of notes wrong or misses them entirely while my ear can discern them clearly. And I'm nowhere close to having a Golden Ear.

 

[elitebasstabs]

I've been testing Elite Bass Tabs and have every hope it turns out well because I really, really want an AI bass tab generator that works, but it's surprising how much better the human ear is over AI still. With isolated bass tracks it still gets a lot of notes wrong or misses them entirely while my ear can discern them clearly. And I'm nowhere close to having a Golden Ear.

You're not wrong, and I appreciate you testing it and telling me directly.

One thing worth checking first, though: using my service, you have to check a box that says "Skip stem separation" if your uploaded audio is already an isolated bass track. I can't see on my end whether you did this. If you uploaded an already-isolated bass track without checking that box, my pipeline will run its own isolation on top of an already-isolated signal — and that can hurt accuracy. The processing is tuned assuming it's working on a full mix; I can't guarantee it behaves the same way on audio that's already isolated, and it could be unknowingly dragging your accuracy down. Worth double-checking that box was checked before we go further. And even if it was checked, I believe — though I'm honestly not sure — accuracy could still take a hit from that isolated track alone. My pipeline's own isolation has specific tuning applied before and after it runs, and I have no way to know the quality or method behind whatever isolation produced your file, so it may not behave the same way through the rest of my pipeline as audio isolated by my own process would. I think the best bet is to run the full mix through my software and let the isolation that the rest of the process is tuned for do its thing.

This is also basically the same reason I started this project. AI doesn't have ears, unfortunately — but it can create and edit code at speeds no human can match. Humans and AI working together, each playing to their own strength, is where I think the future is. I personally have trouble hearing the pitch of lower notes on bass, and wanted a way to learn and transcribe songs faster. AI can't actually hear a note, but pitch-detection software can tell you how confident it is that a note is a certain frequency, and it does that incredibly fast. The problem is there are a lot of factors that can erode that confidence for any single note — that's where the current technology falls short, and until someone makes a real breakthrough there, software transcription is never going to hit 100%.

So why am I still putting so much time into this? I still think it's useful. Even if the project completely fails and nobody ends up using it, I'd still call it a personal success for a lot of reasons. I use it every day and find it extremely helpful and time saving. Of every other AI bass tab service or software I've paid for, all but maybe one were absolute garbage — getting any notes right was optional, and fixing them by hand cost more time than it saved. I also wanted playback that syncs the original audio to the generated tab as you listen — that's been the most useful feature, by far and is something EBT does well. When I factor the speed and accuracy of the generated tab, the player and the fact that the player is also an editor capable of making changes to the notes in the generated tab without affecting the audio/tab sync this is where the usefulness is highest in my opinion. I will admit, the learning curve for the editor is probably kind of high for most, but when I can get it to it I will be creating some extensive how to docs for it and even some videos. But this is later on.

 

[SteveCS]

But even with my 45+ years of experience playing bass I have yet to create a fingering profile that gets it all right.

At the point of performance, everything is the responsibility of the player, and more so if they have not validated the transcription/chart for themselves. I would add an explicit disclaimer/caveat to the output file, something like "Any fingerings given are suggestive and other fingerings may be possible." Basically, it is down to the player to execute as they see fit. Getting it right is a noble mission and I empathise with your challenge, but in the end it isn't your responsibility. That resides with the player, and it is one they/we cannot abdicate.

YMMV