How do reads and writes of Digitakt II projects over USB work (+Drive file API)?
Request types (file/data layer, names as used by our tools)
- 0x53 list · 0x54 read open · 0x55 read partial · 0x56 read close
- 0x57 write open · 0x58 write partial · 0x59 write close
- 0x5A move · 0x5B copy · 0x5C clear · 0x5D swap
This is a separate request layer from the SysEx dump types; some byte values overlap (0x54, 0x58) and mean different things there. Projects live at /projects/1 … /projects/128; /projects/0 answered "Active project not supported".
Write sequence (hardware-verified)
WriteOpen 0x57: u32be total_upload_bytes + path + 0x00 -> reply: file handle (fd)
Write 0x58: u32be fd, u32be seq, u32be crc, u32be chunk_len, data (chunks of 8192 bytes)
seq = 0,1,2,... (chunk index, NOT byte offset)
reply: fd, seq, cumulative bytes accepted
WriteClose 0x59: u32be fd, u32be full_upload_size
- Sending a byte offset (16384) as the second sequence value was refused with "Invalid sequence number".
- Using 1 instead of the full size in WriteClose gave "Footer was not processed".
- One project upload of 891,375 bytes took 109 chunks and was acknowledged in full.
Project file forms
- Upload form: 31-byte header starting
AC 11 D3 03, chained raw-LZ4 blocks of a 14,521,860-byte project image, a zero block terminator, then a 12-byte trailer: CRC32 (zlib-compatible) over the compressed blocks incl. terminator, compressed length, and the markerAA A1 DA AA. Header bytes 29–30 were01 0Cin compressed exports. - Read form: 31-byte header (bytes 29–30
00 0C) + the 14,521,860-byte uncompressed image + 12-byte trailer = 14,521,903 bytes. - So never byte-compare an upload with a readback; compare the decompressed images.
A whole-project SysEx dump does not contain the full project image (e.g. sample-pool metadata), so building an uploadable project needs a compatible base project file.
What we did not find
On stock OS 1.16/1.17 we found no remote command that makes a +Drive project the active project; loading stays a front-panel action. Writing a slot and restoring a SysEx dump into the working buffer are separate steps.
We do not build, distribute or describe custom firmware, and we do not publish anything about firmware encryption. Everything here is ordinary data exchange with the stock OS.
FAQ
Is the write sequence number a byte offset?
No. It is a zero-based chunk index; byte offsets were refused.
Why is my Digitakt II project readback bigger than my upload?
Uploads are LZ4-compressed; reads return the expanded 14,521,903-byte native form.
Is there custom firmware for the Digitakt II?
Not from us. These notes only cover the stock OS's own data exchange.
Observed on / Status: Hardware-verified on Digitakt II OS 1.16 build 0079 (October 2026) for writes to an empty slot; request names partly inferred. Unverified on later OS versions.
Deutsch
Neben SysEx kennt die DT2 dateiartige Anfragen (List, Read, Write, Move, Copy, Clear, Swap). Geschrieben wird in 8.192-Byte-Blöcken mit nullbasierter Sequenznummer und CRC je Block. Hochgeladen wird komprimiert (LZ4), gelesen wird das ausgepackte Format. Keine Custom-Firmware, nichts zur Firmware-Verschlüsselung.
Related
- What does a Digitakt II SysEx message look like?
- How can AI agents jam on a hardware sampler without wrecking the set (clipping, level locks, filter checks)?
- What is an MCP server for music hardware, and how should it be designed?