Skip to content

RØDECaster Duo support ​

RC2FX has experimental support for the Duo on official Duo firmware 1.7.6 and 1.7.7. It is off by default. To flash a Duo, the installer needs a saved opt-in plus a risk confirmation on every run.

None of it has been tried on a real Duo: flashing, recovery and audio are all untested. Everything on this page comes from the official firmware files, static analysis and ARM64 emulation. Flashing a Duo may leave it unusable, recovery may fail, and audio may be loud or wrong. Use it at your own risk.

Do not flash Pro II firmware onto a Duo or Duo firmware onto a Pro II. The two share a mixer program, but their kernel, boot files and controller firmware differ.

Enabling it ​

When the installer is idle, it shows an Allow experimental Duo flashing switch above Developer options. The setting is stored as experimentalDuoFlash in the app's view.json. It defaults to off, and a missing or malformed value is treated as off (crates/rc2fx-app/src/connection.rs).

With the switch on and a Duo connected:

  • Each run, and each retry before the flash step, shows DUO_WARNING (crates/rc2fx-installer/src/lib.rs) and asks for Flash Duo at my own risk. The confirmation is cancelled if the model, serial or release changes while the question is open.
  • For RC2FX, the installer offers only the two pinned official Duo archives, 1.7.7 and 1.7.6. Stock firmware comes from RØDE's Duo manifest, rcp2-mini-manifest.json. It never offers Pro II images for a Duo.
  • All other checks still apply: archive pins and model, board identity, instruction signatures and serial continuity.
  • The prepared image gets a file /opt/rc2fx/experimental-duo containing 1. The launcher (xtask/mixer/rc2fx.sh) turns it into RC2FX_EXPERIMENTAL_DUO=1, which lets the runtime accept this Duo board with the Duo profile. The Duo is still identified as a Duo.

Turning the switch off blocks further Duo flashes. It does not remove RC2FX from a Duo that already has it. To remove it, flash stock Duo firmware, which in the RC2FX installer also requires the switch to be on. The app connects to an installed Duo runtime through the RC2FX USB descriptor and handshake, as it does for a Pro II.

Processing paths ​

The Duo has its own route layout, duo-23-v1, and saved-show source enum, duo-sources-v1. As on the Pro II, processing starts off.

InputsOutputs
Mic 1, Mic 2, Headset micHeadphones 1, Headphones 2, Headset
USB 1, USB 2, ChatMonitors
Game, Music, Virtual A, Virtual B
Bluetooth, CallMe 1-3Main Mix Out, Chat Mix Out, Secondary Mix Out (to USB 1, Chat and USB 2)
Wireless 1, Wireless 2

Every saved Duo target ID starts with duo-, so a Pro II preset cannot put a chain on a Duo socket that is wired differently. The layout includes the four virtual USB inputs. It does not include SMART Pads or the extra per-channel multitrack output feeds. Source names are established separately from route availability, in the source-name evidence.

The Duo routing note gives the port for each path, the firmware addresses behind it and the evidence. In summary:

  • Headphones 1 is out_3/out_4 and Headphones 2 is out_1/out_2. The headset output mirrors Headphones 1 on out_8 (left) and out_7 (right).
  • The USB sends are named connections from mixer ports to USB endpoints: out_15/16 to USB 1, 17/18 to Chat and 19/20 to USB 2. Running the firmware's own connection code confirmed them.
  • Monitors is out_5/out_6. That is the firmware's mapping when all eight hardware outputs are active. Nobody has seen a real Duo's JACK graph at startup or after a reconnect, so this mapping is unconfirmed for some USB and startup modes.

The native pipeline binds ports by name and does not fall back to numbered buffer indexes. A stereo output is processed only when both of its named buffers appear in the vendor callback. Otherwise it is left stock, which prevents processing the wrong buffer or only one channel. The pipeline does not check which physical socket a port is wired to, and having both names present does not prove the stereo mix is correct.

Firmware ​

ReleaseArchive SHA-256Mixer program
1.7.647d2ccb0d613b223ca6d5ed57c62ae6ee32ba71b662be5fbfb4ad3f44c757daaSame as Pro II 1.7.6
1.7.74f3c7a022f07fd5b05955846e6a9052ad6ae214e12c04cda2d4749d775a304edSame as Pro II 1.7.7

Both releases use the same launcher. The hook audit (rodecaster-duo-1.7.7.json) and the route audit (rodecaster-duo-routes.json) cover both, and the routing note explains what each one shows. Both audits were run on the 1.7.7 beta archive. The stable 1.7.7 archive, downloaded again on 2026-11-03, is byte-identical to it.

Identifying the model ​

The mixer program is the same on both models, so its hash cannot identify the mixer. RC2FX describes each model separately (Model in crates/rc2fx-firmware/src/model.rs) and identifies the board directly:

Pro IIDuo
Hostnamerodecaster-pro-iirodecaster-pro-ii-mini
Boot device treemeson-axg-rodecaster-pro-ii.dtbmeson-axg-rodecaster-pro-ii-mini.dtb
Firmware folderrcp2/rcp2-duo/
USB IDs, normal modes0037 0078 008d 0092 00940050 0079 008e 0093 0095
USB ID, update mode0030004f

On the mixer, RC2FX reads the model from /etc/hostname and from the device tree named in /boot/extlinux/extlinux.conf. The two must agree, and there is no override. The USB IDs come from RØDE's gadget script, which both firmwares share. They have not been observed on a real Duo.

How RØDE's app updates a Duo ​

This comes from static analysis of the RØDECaster App 2.0.115 for macOS, whose binary keeps its symbols. The Windows build names the same manifests and files.

  • One class, Rodecaster2FirmwareUpdater, updates the Pro II, the Duo, the Streamer X and an unreleased "RØDECaster Pro II-A". It has no model-specific branch.
  • It sends HID M (report 1, 63 bytes) and expects an A reply. It waits up to 30 seconds for update mode, copies archive.md5 and then archive.tar.gz to the update drive, runs sync, waits 1 second, ejects the drive and sends HID U. A factory reset sends R before M. A mixer that is already in update mode is asked for its drive's volume ID with HID V. On macOS the drive is the mounted NO NAME volume on the same USB device.
  • Each model has its own manifest: manifest.json for the Pro II (0030), rcp2-mini-manifest.json for the Duo (004f), and rcp2-a-manifest.json for the Pro II-A (009f), which is not published yet.
  • RØDE's device table lists the Duo as 004f 0050 0073 0079 00930095, the Pro II as 0030 0037 0072 0078 0092 0094, and the Pro II-A as 009b to 009f. RC2FX does not list 0072 or 0073, and RØDE does not list 008d or 008e. RC2FX ignores a device it does not list.

RC2FX's flash takes the same steps. It also reads both files back before ejecting and waits for the drive to go. None of this tests the RC2FX image or runtime on Duo hardware.

Untested ​

These all need a real Duo, and a separate decision to flash one:

  1. Flashing either release, the first-boot controller update, and returning to stock.
  2. USB mode changes, reconnects and the app's control session.
  3. Each path with an audio loopback, including Monitors in each USB mode and the headset's channel order.
  4. Source reassignment, fallback to stock on failure, and sustained processing load. The figures in Hardware are for the Pro II.

For a first test, back up the RØDE show, disconnect headphones and speakers, and leave the mixer alone through its first boot after the flash.

Offline development ​

This work needs no Duo. The Duo firmware is not in git. Download https://update.rode.com/rcp2-duo/1.7.6/os.tar.gz and https://update.rode.com/rcp2-duo/1.7.7/os.tar.gz to .deps/duo-investigation/duo-1.7.6-os.tar.gz and duo-1.7.7-os.tar.gz.

To check an official archive without extracting it:

sh
for v in 1.7.6 1.7.7; do
  cargo run -p rc2fx-installer --example inspect-firmware -- \
    rodecaster-duo ".deps/duo-investigation/duo-$v-os.tar.gz"
done

This verifies the checksums, that the root filesystem and boot files are a Duo's, the mixer and launcher hashes, and the bytes RC2FX hooks. A good result shows executableVerified: true. runtimeQualified stays false, because the Duo runtime is accepted only with the opt-in.

To check an extracted mixer and launcher:

sh
cargo run -p rc2fx-firmware -- inspect config/firmware-profiles.json \
  .deps/duo-investigation/duo/usr/bin/rc_audio_mixer \
  .deps/duo-investigation/duo/usr/bin/audio-start.sh rodecaster-duo

Tests that cover Duo handling:

sh
cargo test -p rc2fx-firmware -p rc2fx-installer -p rc2fx-device
cargo test --manifest-path crates/rc2fx-native/Cargo.toml --lib
cargo test -p rc2fx-app installer::
python3 xtask/mixer/tests/usb_restart.py

# These need the official archives:
for v in 1.7.6 1.7.7; do
  RC2FX_DUO_ARCHIVE="$PWD/.deps/duo-investigation/duo-$v-os.tar.gz" \
    cargo test -p rc2fx-installer pinned_duo_archive -- --ignored
done

To build a temporary Duo image from an official archive and check that it keeps the Duo's boot and controller files and has the opt-in marker, first cross-build the mixer runtime locally:

sh
cargo xtask mixer-build
RC2FX_DUO_ARCHIVE="$PWD/.deps/duo-investigation/duo-1.7.7-os.tar.gz" \
  RC2FX_REAL_ROOT="$PWD" \
  cargo test -p rc2fx-installer real_duo_image -- --ignored

The routing note has the audit commands and the offline results so far, under its verification receipt. None of those results involve a real Duo.

Factory self-test mode ​

RØDE's launcher has a "production" branch, selected by the file /usr/bin/production-boot. It is the factory self-test mode. It runs at 128 frames, reflashes the VL805 controller firmware, deletes saved system-config* files, changes bootloader settings and starts a self-test engine. Do not create that marker, including to experiment with buffer sizes. RC2FX supports only normal-mode 112-frame operation on both models.