TL;DR
An open-source project called CoyoPedal runs full-size Neural Amp Modeler A2 captures in real time on the Waveshare ESP32-S3-Touch-AMOLED-2.06 development board, using it as a standalone guitar amp and effects pedal. The firmware also compiles to WebAssembly for browser testing without hardware.
An open-source project has turned the Waveshare ESP32-S3-Touch-AMOLED-2.06 development board into a standalone guitar amp and effects pedal that runs full-size Neural Amp Modeler A2 captures in real time at 48 kHz. According to the project’s GitHub repository, the device — called CoyoPedal — also acts as a USB host for a class-compliant audio interface and offers a touchscreen interface, with the same firmware compilable to WebAssembly so it can be tried in a browser without any hardware.
Get audio and creator gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The project, published at github.com/dashersw/coyopedal, states that the amp modeling is handled by a 23-layer, eight-channel WaveNet running in block floating point with hand-written Xtensa kernels, split across both of the ESP32-S3’s cores and processed in 64-frame blocks. The repository reports that on a board with no screen, the DSP load was measured at 91% and 94% of the 1,333 µs block budget across the two cores, with no missed deadlines. The A2 model lives in the board’s 8 MB of PSRAM, which the project lists as a hard requirement alongside a USB port the chip can drive as a host.
Audio enters and leaves through a USB Audio Class 2 interface attached to the board’s USB port, which runs as a USB host at 48 kHz. The project says it has been tested with the XTONE Pro and IK Multimedia iRig HD X, with the iRig HD 2 supported through a dedicated UAC1 profile. Because interfaces are discovered from their USB descriptors, the project states that any interface exposing a 48 kHz UAC2 input and output should work, though this is a general claim rather than a per-device guarantee.
Users can load their own captures by copying original .nam files to a microSD card (or pointing the browser build at a folder). The pedal parses, validates and prepares the models on the device — the project stresses that there is no desktop converter. Supported models are 48 kHz, eight-channel NAM A2 (“A2-Full”) WaveNets; other architectures, sample rates and layer shapes are rejected with an error. The first selection of a capture pauses audio while the pedal prepares it, which the documentation says can take tens of seconds; a verified .s3cache file is stored next to the original for faster later loads, and the original file is never modified.
Why a $30 Microcontroller Changes Amp Modeling
Neural amp modeling has until now been largely the domain of desktop software, dedicated hardware units like the Neural DSP Quad Cortex, or single-board computers. Running a full-size NAM A2 capture on an ESP32-S3 — a microcontroller typically costing a fraction of those platforms, with fixed-function peripherals rather than an operating system — shows how far hand-optimized fixed-point-style DSP can stretch modest hardware. For guitarists, it points toward cheap, hackable, self-contained amp modeling that does not depend on a laptop or a proprietary ecosystem.
The on-device capture preparation is a second notable departure from common practice: most NAM workflows require a desktop tool to convert captures before deployment. The project also publishes a browser build at coyopedal.playtaurus.com that runs the same UI and the same DSP compiled to WebAssembly, removing the hardware barrier to trying the software entirely.
Products Worth Considering
Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity, Black
Bose QuietComfort Ultra Wireless Over-Ear Headphones with Spatial Audio and Noise Cancellation, Black
Kindle Paperwhite Signature Edition 32GB E-Reader, 7-Inch Display, Seaglass Green
As an affiliate, we earn on qualifying purchases.
The Boards, Effects Chain and Presets
The reference hardware is the Waveshare ESP32-S3-Touch-AMOLED-2.06, which carries an ESP32-S3R8 chip, 32 MB of flash, 8 MB of PSRAM, a 410 × 502 AMOLED touchscreen, an AXP2101 power-management chip and a microSD slot. The same firmware also builds for the ESP32-S3-DevKitC-1 N16R8 and bare S3R8 modules with no panel at all — the project notes that on a bare module the BOOT button acts as the footswitch, and configuration is done over a maintenance API rather than by hand. A bare module has no power-management chip or battery path, so the USB audio interface needs external power.
Signal processing is not limited to the amp. The chain includes a gate, compressor, chorus and drive before the amp block, and a digital delay and stereo spring reverb after it. Presets for amp, controls and effects are stored on the pedal itself, with an on-screen keyboard for naming, and a tuner shares the screen with the output muted while tuning. A maintenance mode provides authenticated Wi-Fi over-the-air updates, remote diagnostics and BLE discovery; the project states the radios are completely off while playing.
The full VoLum library of captures is published in a separate repository and can be copied to a card with a single npm command. The browser build asks for a folder instead of a card and answers the firmware’s SD-card calls from it, showing captures under the same names and IDs they would have on hardware.
“The 23-layer, eight-channel WaveNet runs in block floating point with hand-written Xtensa kernels, split across both cores and processed in 64-frame blocks.”
— CoyoPedal GitHub repository
Claims That Await Independent Testing
All performance figures — including the 91%/94% core-utilization measurements and the claim of no missed deadlines — come from the project’s own documentation and have not been independently verified. The statement that “any interface that exposes a 48 kHz UAC2 input and output should work” is the project’s expectation based on descriptor-based discovery, not a tested guarantee across devices.
The repository does not state the project’s licensing terms in the material provided, its roadmap, or how the tone of the prepared models compares subjectively to desktop NAM implementations. Details on pricing, availability of prebuilt hardware, or any commercial plans are also absent — this appears to be a self-published open-source effort, and it is not yet clear how widely it has been adopted or reviewed by the guitar community.
Trying It: Browser First, Board Second
Interested readers can try the firmware without hardware at coyopedal.playtaurus.com, using a guitar and a USB audio interface with nothing to install. To build the physical pedal, the project’s documentation points to the Waveshare AMOLED board as the reference platform, with the ESP32-S3-DevKitC-1 as a screenless alternative. Future developments to watch include community testing of the performance claims, broader USB interface compatibility reports, and any expansion of supported capture architectures beyond 48 kHz eight-channel A2 WaveNets.
Key Questions
Do I need to buy the Waveshare board to try CoyoPedal?
No. The same firmware compiles to WebAssembly and runs at coyopedal.playtaurus.com in a browser, using the same UI, DSP and neural amp model. You need a guitar and a USB audio interface, but no board and nothing to install.
Can I use my own Neural Amp Modeler captures?
Yes, if they are 48 kHz, eight-channel NAM A2 (“A2-Full”) WaveNets, including matching members of a SlimmableContainer. Copy the original .nam or prepared .namb files to a microSD card under a nam folder, or point the browser build at a folder. Other architectures and sample rates are rejected with an error.
Which USB audio interfaces work with it?
The project has tested the XTONE Pro and IK Multimedia iRig HD X, with the iRig HD 2 supported through a dedicated UAC1 profile. Interfaces are discovered from their USB descriptors, so the project states any 48 kHz UAC2 interface should work, though this is unverified for specific devices.
Can the pedal run on a cheaper board without a screen?
Yes. The same firmware builds for the ESP32-S3-DevKitC-1 N16R8 and bare S3R8 modules. The amp, effects and presets are identical, the BOOT button acts as the footswitch, and the device is configured through maintenance-mode tooling. The board must have 8 MB of PSRAM and a host-capable USB port; a bare module cannot power the USB interface itself.
How long does it take to load a new capture?
According to the project, the first time a capture is selected, audio pauses while the pedal prepares it, which can take tens of seconds. A verified .s3cache file is stored next to the original file when the card is writable, so later loads are faster; the original file is never modified.
Source: hn
