[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86463-en":3,"doc-seo-86463-105":29,"detail-sidebar-cat-0-en-105":82},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},86463,8796095461564,"Liam","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",6,"Technology","WaveNet-Style Guitar Amplifier Model Pruning for Real-Time iOS Deployment","WaveNet-style convolutional networks emulate tube amplifiers and distortion pedals with high-fidelity audio, but their computation has limited use to desktops or dedicated DSP hardware. A sparse-enabled WaveNet inference engine for iOS enables real-time execution of heavily pruned neural guitar amplifier models on CPU-only iPhones. Iterative magnitude pruning removes 90% of weights without perceptible quality loss, and a custom sparse C++ engine converts sparsity into measurable compute savings. On-device audio matches the trained model within int16 quantization error, demonstrated via A/B comparison against the emulated physical pedal.","Proceedings of the 29th International Conference on Digital Audio Effects (DAFx26), Cambridge, MA, USA, 1 –4 September 2026 (Demo)  \nWAVENET-STYLE GUITAR AMPLIFIER MODEL PRUNING FOR REAL-TIME IOS  \nDEPLOYMENT  \nRyota Sato and Eli Silverstein  \nDept. of Electrical Engineering  \nStanford University  \nStanford, USA  \n[ryos17@stanford.edu](ryos17@stanford.edu)  \n[esilvers@stanford.edu](esilvers@stanford.edu)  \narXiv :2607 . 10086v1 [ ee ss .AS] 11 Jul 2026  \nABSTRACT  \nWaveNet-style convolutional networks emulate tube amplifiers and distortion pedals with high fidelity, but their computational cost has confined them to desktops or dedicated DSP hardware. We present a sparse-enabled WaveNet inference engine for iOS that runs heavily pruned neural guitar amplifier models in real time oniPhones. Aggressive iterative magnitude pruning removes 90% of the network weights with no perceptible loss in quality. A custom sparse C++ engine turns this sparsity directly into compute savings, sustaining low-latency real-time operation on a CPU-only iPhone implementation where the dense model cannot. On-device output matches the trained model to within int16 quantization error. At the demonstration, visitors will play a guitar through the app on iPhone hardware and A/B the on-device pruned model against the physical pedal it emulates. Source code and audio examples are available at [https://github.com/ryos17/wavenet-imp](https://github.com/ryos17/wavenet-imp).  \n1. INTRODUCTION  \nVirtual analog (VA) modeling of audio circuits has become an active research area, particularly for guitar amplifiers and effects [1] . White-box methods explicitly simulate the underlying circuit [2] but require detailed component knowledge, whereas black-box methods learn the nonlinear input–output relationship directly from measured data. Neural networks are especially effective here, with early approaches using Long Short-Term Memory (LSTM) recurrent networks [3, 4] and state-of-the-art methods using WaveNet-style convolutional networks with dilated one-dimensional convolutions [5] .  \nSuch models have traditionally been deployed on desktop machines or dedicated hardware [6] because of their high computational cost. To improve efficiency, prior work has turned to iterative magnitude pruning [7], which removes a large fraction of weights while preserving accuracy, as shown for LSTM-based neural amp models [8] . Such sparse subnetworks are supported by the lottery ticket hypothesis [9] and confirmed for convolutional audio models by Esling et al. [10] . However, systematic pruning of WaveNetstyle neural amplifiers remains largely unexplored, and little effort has targeted sparsity-aware inference engines for neural amplifier modeling on portable consumer devices such as the iPhone.  \nThis work demonstrates the real-time deployment of a heavily pruned WaveNet-style network for guitar amplifier emulation on a  \nCopyright: © 2026 Ryota Sato et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4 .0 International License, which permits unrestricted use, distribution, adaptation, and reproduction in any medium, provided the original author and source are credited.  \nFigure 1: WaveNet-style architecture used for neural guitar amplifier modeling, reproduced from [5] .  \nCPU-only iOS implementation. Section 2 details the training and pruning strategy, and Section 3 the real-time iOS C++ inference engine and its on-device performance. At the demonstration, attendees will play through the on-device model and compare it against the original analog hardware.  \n2. MODEL TRAINING AND PRUNING  \n2.1. Architecture  \nWe adopted a variant of the original WaveNet architecture [11], a causal feedforward network composed of stacked dilated convolutional residual blocks (Figure 1) . The network learned a direct mapping from the raw input guitar waveform x[n] to the corresponding distorted output yˆ[n] . Dilation expanded the receptive field of ","cbCaicLBZP4VmR9R","https://ap.wps.com/l/cbCaicLBZP4VmR9R","pdf",2841829,1,4,"English","en",105,"# 1. INTRODUCTION\n# 2. MODEL TRAINING AND PRUNING\n## 2.1. Architecture\n## 2.2. Training Objective","[{\"question\":\"How is on-device output quality evaluated in the demo?\",\"answer\":\"Visitors can play a guitar through the app on iPhone hardware and A/B the on-device pruned model against the analog pedal being emulated, while the document also states the output matches the trained model within int16 quantization error.\"}]",1784211880,10,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":77,"head_meta":79,"extra_data":81,"updated_unix":27},"wavenet-style-guitar-amplifier-model-pruning-for-real-time-ios-deployment","",{"@graph":35,"@context":76},[36,52,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/technology/",3,{"item":51,"name":13,"@type":42,"position":21},"https://docshare.wps.com/document/wavenet-style-guitar-amplifier-model-pruning-for-real-time-ios-deployment/86463/",{"url":51,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":23,"description":14,"dateModified":60,"datePublished":61,"encodingFormat":59,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":40,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70],{"name":71,"@type":72,"acceptedAnswer":73},"How is on-device output quality evaluated in the demo?","Question",{"text":74,"@type":75},"Visitors can play a guitar through the app on iPhone hardware and A/B the on-device pruned model against the analog pedal being emulated, while the document also states the output matches the trained model within int16 quantization error.","Answer","https://schema.org",{"og:url":51,"og:type":78,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":80,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":83},[84,88,92,96,101,104,109,114,119,122,125],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":102,"slug":103},50,"technology",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},7,"Healthcare",40,"healthcare",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},8,"Research & Report",30,"research-report",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":45,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":28,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":28,"slug":124},"Lifestyle","lifestyle",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":97,"slug":128},19,"General","general"]