Virtual MPC Sample

English · Русский · 中文

The stock firmware of the Akai MPC Sample (1.3.0.12) running unmodified in QEMU, with the whole device in your browser: the top panel with the live screen, pads, knobs and buttons, the rear panel with its jacks, the SD slot and USB-C. Works on macOS, Linux and Windows; fastest on Apple Silicon.

Factory library on /content	Trim page	Power key: the app asks	...saves and powers off

Unofficial project. Not affiliated with, endorsed by or supported by inMusic Brands or Akai Professional. No firmware and no factory content are in this repository: you need your own copy of the firmware update and, for the factory library, your own MPC Sample.

What it is for

• Build and test mods without risking the device. SD autorun packages run exactly as on the MPC (the firmware's own script runner); a broken package costs a lab.py reset, not a recovery.
• Debug inside the stock app. Root shell, LD_PRELOAD libraries, strace, the panel's raw control-surface bytes.
• Learn the device. stock-map/ is every stock screen, in English, Russian and Chinese, with what each button and SHIFT combination does there.

The stock screen map: 56 sections, 383 screenshots taken in this emulator, one offline HTML page (open stock-map/index.html).

What works

• The stock rootfs, byte-identical, opened through dm-verity with the stock root hash from the signed kernel image; the stock systemd units start the stock app.
• The 320×240 screen, pads, knobs, buttons and the data wheel through the same UART protocol the panel's microcontroller speaks.
• Sound in the browser page; sampling from the page's inputs.
• SD card (any folder on your computer), SD autorun packages, the factory library.
• USB-C: the device's own USB network (192.168.155.1), USB MIDI, root SSH for community packages.
• Power key as on the device: the app asks, saves the project, powers off.

Not yet: USB audio over USB-C, a Windows test run, other firmware versions. See the roadmap.

Quick start

You need QEMU 9+, Docker (only for the build), Python 3.9+ and an OpenSSH client.

1. Clone the repository.

2. Put the firmware partitions in firmware/extracted/ (git-ignored): part1 (boot), part2 (rootfs), parr (recovery), extracted from inMusic's MPCSAMPLE-1.3.0.12-update.img. The build checks them against inputs.sha256. This repository has no extractor for the update image yet (details).

3. Build and run:

   python3 lab.py build    # Docker: kernel (~15 min once), audio bridge, disk images
   python3 lab.py run      # the VM; starts the panel too
   

4. Open http://127.0.0.1:8765/ and click once for sound.

python3 lab.py ssh               # root shell in the VM
python3 lab.py autorun PKG.zip   # run an SD autorun package
python3 lab.py stop              # power key + "Yes": saves, shuts down, QEMU exits
python3 lab.py reset             # next start from a clean eMMC

On Windows use py lab.py .... On macOS start the lab from Terminal.app or iTerm, not from an IDE's or another app's terminal pane: macOS may throttle it, and the sound clicks.

Documentation

docs/USAGE.md	full guide: factory library, USB-C, SD autoruns, host requirements, ports
docs/PANEL.md	the virtual panel, its protocol, sound, the SD card
docs/CUSTOM_FIRMWARE.md	your own code: autorun packages, LD_PRELOAD, lab layer files, other firmware versions
docs/MPC_SAMPLE_VIRTUAL_LAB_ARCHITECTURE.md	what is known about the device and why the lab is built this way
docs/POC_RESULTS.md	measurements and the differences from the device that remain
stock-map/	every stock screen, EN/RU/中文, one offline HTML page

Folder	Contents
lab.py	the whole lab: build, run, ssh, autorun, stop, selftest
panel/	the browser panel and its server (plays the panel MCU)
kernel/	lab kernel config and patches (mainline 6.12.60 PREEMPT_RT)
overlay/	the lab-only read-only layer over the stock rootfs
bridge/, initramfs/, docker/	audio bridge, lab initramfs, build container
scripts/, tools/	build scripts and host tools

Roadmap

Help is welcome on any of these:

• Firmware extractor for the AZ0x update image, so that setup is one command. The biggest barrier for new users today.
• Windows and Linux runs. The code paths exist; only macOS (Apple Silicon) has been run.
• Prebuilt kernel and audio bridge as release downloads, so no 15-minute Docker build.
• CI running make test (the lab and panel self-tests) on every pull request.
• USB audio (UAC2) on the emulated USB-C; it is left out because it cannot bind on the lab's USB controller.
• Fewer audio clicks under software emulation (TCG) and a tickless guest under HVF without the nohz=off cost.
• Other firmware versions and checks of what each one changes.
• Host MIDI in the panel page (Web MIDI to the rear MIDI jacks).

Open an issue before a large change. See CONTRIBUTING.md.

Built with Claude

This project was developed together with Claude, Anthropic's AI model, in Claude Code. Claude wrote much of the code, the kernel patches and the documentation, and ran most of the measurements in the VM. The author set the goals, made the decisions, checked the results against a real MPC Sample and is responsible for what is published here.

Treat it as you would any young project: read what you run. The figures in the docs were measured on one Mac (Apple Silicon, firmware 1.3.0.12) on the dates given.

Legal

• Akai Professional and MPC are trademarks of inMusic Brands, Inc. They are used here only to name the device the lab emulates.
• The lab only reads the firmware files you provide, mounts them read-only and never writes to a physical device.
• Screenshots of the stock app are shown for documentation and interoperability.

License

MIT, except kernel/patches/, which are patches to the Linux kernel and are under GPL-2.0, as the kernel is.