Sebastian Murgul

CEO at Klangio

Sebastian Murgul is the co-founder and CEO of Klangio GmbH, a company that develops apps for automated transcription of music. He studied Electrical and Information Engineering at the Karlsruhe Institute of Technology (KIT) where he earned his Master of Science. As an external doctoral candidate at KIT, he has continued to pursue his research interests in Music Information Retrieval (MIR) since 2021. His research focuses on the automated transcription of guitar music signals using neural networks, up to quantized tablature.

🌐 Speaker webpage

From audio signal to sheet music with AI

On April 24, 2024 at 05:00 PM (UTC+1)

In his presentation, Sebastian Murgul introduces the innovative technology of the startup Klangio in the field of automated music transcription, with a focus on the application of artificial intelligence to convert music into notation. The lecture begins with the basics of music transcription, followed by a brief introduction to the topic of neural networks. After presenting the technologies behind Klangio, practical use cases are shown in a live demonstration. The presentation concludes with a look at the future of music transcription. The aim of the lecture is to provide insights into the possibilities of AI-supported transcription for musicians and the music industry and to demonstrate its potential.

Ask your questions to Sebastian Murgul!

How can AI-supported music transcription by Klangio change the work of musicians and the music industry?

by Charlotte Bettermann

Innovative technology is mentioned and so I am curious - what does Klangio do differently from, for example, AnthemScore's transcription with convolutional neural networks (2016)? This task seems more mathematical than anything, and so I wonder just how of a difference smaller machine learning improvements might be making.

by Anastasia Shulman

What are specific challenges in developing AI algorithms for the transcription of music, for example to capture the nuances of guitar music?

by Anna-Lena Scheunemann

What are the next steps for Klangio in terms of technological advancements or new features that you are most excited about implementing?

by Kathrin Liesens

I know from experience that sheet music can be very pricey. Therefore I could imagine someone trying to reproduce the sheets for copyrighted songs for a very low or no price at all by using Klangio (or the free trial available).
So can the used AI in Klangio identify songs with copyright? Can it block the output?

by Mercedes Neuwirth

How does Klangio ensure that its technology remains compatible with new or existing music production software?

by Melanie Bauer

Today the tool is available for piano, guitar, bass, and vocal recordings. Is imaginable to develop Klangio for every instrument and at the end transcript orchestral audio files (with all the instruments separately)? What does it require technically for the AI model?

by Antonin Couton

According to the app Sing2Notes, how does AI solve the problem of people not being able to sing as precisely as musical instruments?

by LiHsuan Chang

In terms of data preprocessing, what steps does Klangio undertake to enhance the quality of input audio for transcription? Are there any specialized techniques employed to handle noise, varying recording qualities, or other audio artifacts?

by Janis Reisenauer

How accurate are the transcriptions provided by guitar2taps, for example, and to what extent do they take into account human error and inaccuracies in timing as well as the clarity and quality of the audio recording?

by Annika Grauer

When instruments are played simultaneously, what challenges does the AI face in such scenarios?

by Niklas Evmenenko

Is it currently possible to transcribe orchestral music, considering the significant variations in volume and the challenge of distinguishing individual instruments?

by Katharina Grünwald

Is there any issue with transcribing pieces regarding Copyright?

by Nathalia Cuenca

What do you consider the key challenge in achieving accurate and reliable transcription of audio music signals into sheet music using AI, and how does Klangio address this issue in its technology and approach?

by Johanna Fritzmeier

Are there any problems or errors in using copyright using AI?

by Junghyun Lee

How does Klangio's technology compare to traditional methods of music transcription and direct competitors in terms of accuracy, efficiency and price?

by Timo Warendorf

Thank you for the interesting lecture. I heard that high-quality audio tracks are needed to obtain the most accurate music transcriptions. But if I transcribe based on low-quality audio tracks, the accuracy might suffer. What can I do to improve accuracy in this case?

by Sumin Lee

Due to the complex nature of the music structure, it is assumed that minor errors may occur in the process of transcription. I wonder if you provide enough friendly interfaces for users to conveniently correct errors that occurred at this time, and what programming you use.

Also, can AI also recognize and process multiple instruments when they are played at the same time?

by Daeun Jeong

How long does it typically take to transcribe a song using Klangio, and what factors can affect this time?

by Vlad Panait

Do you plan to create an AI-System which is able to recognise every instrument without the need to switch to another system like now (Sing2Notes, guitar2taps …)? Maybe also at the same time?

by Samuel Pucher

Considering your expertise, how do you see the role of AI evolving in music education, particularly in teaching complex instruments like the guitar, and how might it impact the way musicians learn and create music in the future?

by Amir Al Nahawandi

I listened to the piano2notes example and wondered what the AI pays attention to. For example, does it also detect the tempo and volume? And how does Klangio react to complex pieces?

by Isabella Deisinger

Is it possible in the future that this AI can analyse an entire song with different instruments and vocals and create the correct sheet music for each?

by Daniel Wiesenfeld

How do the Klangio apps ensure that their note transcription remains accurate while transcribing all sorts of melodies, especially considering the nuances and complexities inherent in different musical genres and styles?

by Michael Buchholz

what potential impacts do you foresee this technology having on musicians and the broader music industry over the next decade?

by Alisa Valentiner

How can neural networks or other AI models contribute to improving the accuracy of converting audio signals into sheet music?

by Sophie Schmitz

I think this program can be very helpful for people. How did you conceive the idea of a program that converts an individual's performance into a notation? Furthermore, what do you believe to be the primary impact or effect of this program?

by Eunwoo Kim

Could you share what your target customer is in the future? I saw many on the website, such as students, professional musicians, etc. Whoever is interested in creating music can use this tool as a helper. For example, in singing, humans' singing tone sometimes can't be the same as that of musical instruments. How did Klangio deal with this problem?

by Kuan Hua Chen

Could you share what your target customer is in the future? I saw many on the website, such as students, professional musicians, etc. Whoever is interested in creating music can use this tool as a helper. For example, in singing, humans' singing tone sometimes can't be the same as that of musical instruments. How did Klangio deal with this problem?

by Kuan Hua Chen

What innovations is Klangio planning in the coming years in the field of sound technology that go beyond what is currently available on the market? Especially if AI continues to develop as is expected?

by Marla Ridzewski

Does Klangio perform equally well across all music genres, or are there specific ones its especially good or bad at?

by Jannik Vieler

Do you have plans to use your technology in other areas outside of music transcription? If yes, which?

by Marco Kosa

Assuming you have a song and want to filter out and transcribe all the different instruments. Could Klangio recognize different instruments as such and write separate transcriptions? Or would the instruments that are played together be mixed into one transcription?

by Lena Ertl

In your journey from studying Electrical and Information Engineering to co-founding Klangio GmbH and delving into Music Information Retrieval research, what pivotal moment or realization sparked your passion for bridging the gap between technology and music?

by Eduard Krasnov

What innovative technologies differentiate the Klangio program from other music AI applications on the market?

by Sarah Obermaier

Is it possible in the future that this AI can analyse an entire song with different instruments and vocals and create the correct sheet music for each?

by Daniel Wiesenfeld

with AIs ever increasing cost of compute and resources do you fear a concentration of power in the big tech companies who can afford such and „outspend“ smaller startups. or do you think startups can be successful even with the lack of resources if so why?

by Julius Greppmair

Thank you for the interesting lecture. I heard that high-quality audio tracks are needed to obtain the most accurate music transcriptions. But if I transcribe based on low-quality audio tracks, the accuracy might suffer. What can I do to improve accuracy in this case?

by Sumin Lee