MENU

Fun & Interesting

Workshop: GPU-Powered Neural Audio - High-Performance Inference for Real-Time Sound Processing - ADC

Video Not Working? Fix It Now

https://audio.dev/ -- @audiodevcon​ --- Workshop: GPU-Powered Neural Audio - High-Performance Inference for Real-Time Sound Processing - Alexander Talashov & Alexander Prokopchuk - ADC 2024 --- Join our GPU AUDIO workshop where you'll dive into the exciting world of neural networks applied to real-time audio processing. In this hands-on workshop, you’ll work with Neural Amp Modeler, an open-source project that uses deep learning techniques to replicate guitar amplifiers and pedals with state-of-the-art accuracy. You will learn how to port and scale the Neural Amp Modeler plugin to the GPU using GPU AUDIO technology stack. With a focus on low-latency, parallel execution, and flexibility, you'll combine different neural building blocks to create high-performance audio models. Whether you're a neural network enthusiast or a DSP pro, this session will showcase how easily scalable models can be built with our specialized GPU neural building blocks. Throughout the workshop, you'll work within a Jupyter environment, building and testing various versions of the Neural Amp Modeler. After the session, you’ll gain access to the workshop’s codebase and environment, allowing you to set it up on your own machine and continue experimenting with real-time, low-latency audio processing. Whether you're working with NVIDIA, AMD, or Mac (M-Silicon) platforms, this session serves as an exclusive sneak peek into our broader release of GPU-powered neural building blocks, unlocking the potential for real-time, scalable, low-latency audio processing like never before. --- Slides: https://data.audio.dev/workshops/2024/gpu-powered-neural-audio-workshop/slides.pdf Chapters: 0:00 Introduction 2:16 Running Neural Amp Modeler using GPU Audio SDK 3:40 Embedded GPUs on NVIDIA Jetson 6:18 GPU Audio Presentation: Neural Amp Modeler 14:02 GPU Audio Supported Platforms 17:15 SDK Workflow Schematics 25:20 Cross Platform Capabilities 26:10 Processor Launcher: Entities 29:05 Processor API 31:10 NAM Models 35:22 Wavenet 36:58 Top Level NAM Core 39:30 Process: Layer Array 40:44 Process: Layer 41:40 GPU Building Blocks Used Today 43:48 Multichannel Delay Line 45:05 Matrix 47:03 Matrix Multiplication 49:18 Conv1x1 52:14 Device Execution: Quick Info 54:38 Performance Info: NVIDIA 4090s 1:00:38 Performance Info: Mac M2 Max 1:05:50 Q&A Session 1 1:22:08 WORKSHOP: GPU Audio SDK 2:10:30 Future & Challenges to Solve 2:13:06 NAM SDK Conversion Overview 2:17:08 Q&A Session 2 2:45:05 Running 100+ NAM Instances on GPU in Reaper --- Alexander Talashov Alexander Prokopchuk --- ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career. Annual ADC Conference - https://audio.dev/ https://www.linkedin.com/company/audiodevcon https://twitter.com/audiodevcon https://facebook.com/audiodevcon https://instagram.com/audiodevcon https://www.reddit.com/r/audiodevcon/ https://mastodon.social/@audiodevcon --- Streamed & Edited by Digital Medium Ltd: https://online.digital-medium.co.uk --- Organized and produced by JUCE: https://juce.com/ --- Special thanks to the ADC24 Team: Sophie Carus Derek Heimlich Andrew Kirk Bobby Lombardi Tom Poole Ralph Richbourg Jim Roper Jonathan Roper Prashant Mishra #adc #audiodev #dsp #audio #gpu #audioprocessing #audioproduction #audioprogramming #neuralampmodeler #neuralnetworks #musictech #soundtech #audiotech #audiotechnology

Comment