This page is read from MIDI Instrument Definitions, a Python library used in the author's other projects, in which every number is traced to a document its maker published. Its source is on GitHub.

Behringer DeepMind 12 MIDI implementation

The MIDI implementation of the Behringer DeepMind 12: what it answers to, and the 269 controls it offers, on firmware 1.1.2.

At a glance

Firmware

This page describes firmware 1.1.2.

MIDI channel

Any one of 1 to 16.

Voices

12.

Note range

Not recorded.

Velocity

Sent and received. Release velocity is received.

Aftertouch

Channel pressure, one value for every key. It sends aftertouch too.

Pitch bend

2 semitones, and the range can be changed.

Program change

Received and sent, with 1024 presets.

MIDI clock

Sent and received.

Start and stop

Not recorded.

Control change

105 controls, in the table below.

NRPN

Supported. 177 of the controls below have one.

System exclusive

Supported.

Controls

In the order the definition gives them, by group. Under each control is the name a program calls it by, once pymidiinstrumentdefs.load("behringer/deepmind_12") has loaded them.

Where these facts come from

Source documents

• User Manual DeepMind 12, 166-page edition of December 2017, its user manual, read on 9 October 2026.
• User Manual DeepMind 12, 135-page edition of January 2017, its user manual, read on 9 October 2026.
• DeepMind 12 Quick Start Guide, V 5.0, its quick start guide, read on 9 October 2026 (the file read).
• DeepMind Family Firmware, 1.1.2, dated 26 February 2018, its release notes, read on 9 October 2026 (the file read).

How they were read

A THIRD-PARTY COPY OF BEHRINGER'S OWN MANUAL, BECAUSE BEHRINGER SERVES NONE. The DeepMind 12's product page lists six quick start guides, the 1.1.2 release notes, four app builds and two preset packs, and no manual. The only link Behringer gives to the User Manual is DeepMind Manual on the DeepMind 12X's page, and that address answered 404 NoSuchKey on 2026-10-07 and again on 2026-10-09. The guide it does serve sends the reader on: "For mapping please consult the product manual." (quick_start p. 34). So the manual cited is Adorama's copy, fetched from Adorama's site: Behringer's own typesetting, 166 pages with a text layer, made in November 2017 and last changed in December. A second, independent copy, Juno's, turned out to be an earlier edition, 135 pages made in January 2017. Both carry the MIDI chapters, and their controller tables and NRPN tables are the same cell for cell, but for the spaces its ligatures leave in the earlier one's text, Eff ects for Effects. Every number the copy shares with what Behringer serves was checked against that, and set out below; if Behringer's own copy returns, it replaces this one.

THE MAP IS TWO TABLES. The controller table, "MIDI Implementation Chart v. 2.0 Control Number Information" (manual p. 133), runs over pp. 133-135 with a row for each of 128 numbers; the NRPN table runs over pp. 136-139. Both were rebuilt from where each cell sits, against the column headings each page prints, because the even pages are set about fifteen points to the right of the odd ones. CC and NRPN are both answered to whatever the instrument is set to send: "Note: Both MIDI and NRPN data is always received and actioned." (manual p. 26). What it sends is a setting per port, Off, Cc or NRPN, where off means "Off (No controller messages will be sent)" (manual p. 26), so every control below travels both ways, which is the format's default and why none says so.

105 CONTROLLERS ARE CARRIED, AND 23 ROWS ARE NOT. Seven are empty, a dash in every column: 3, 9, 14, 15, 22, 118 and 119. Eight are how an NRPN or RPN travels rather than parameters. The NRPN chapter's message is B0, 63, aa, 62, bb, 06, cc, 26, dd (manual p. 136), the parameter number on 99 and 98 and the value on 6 and 38; 96 and 97 are Data Increment and Data Decrement; and 100 and 101 select an RPN, which the chart gives No both ways. And 120 to 127 are the channel mode messages, which this format refuses as controls. Bank select is carried, at 0 and 32, with the banks the chart names: Bank A(LSB =0) to Bank H (LSB =7) (manual p. 132).

177 NRPNS ARE NUMBERED ONE TO A ROW, FROM 0 TO 241, AND FIVE ROWS ARE SPANS. The four effect slots' twelve parameters each, 167-178, 180-191, 193-204 and 206-217, and the program name, 223-239, are each one row over a run of numbers (manual p. 139), and carrying them would mean writing numbers the page does not print one by one, so they are not carried. Every other number from 0 to 241 is.

A CONTROLLER CARRIES AN NRPN ONLY WHERE THE TWO TABLES PRINT ONE NAME, ONCE EACH, as for the Prophets: thirteen times, from LFO 1 Rate at 16 and NRPN 0 to the Mod envelope's sustain curve at 60 and NRPN 78. Two of the thirteen differ only in case, MOD Envelope against Mod Envelope. Every pair's NRPN is 0-255, which is the nrpn_range below; the controller table prints no range at all, so each controller keeps the format's default of 0-127. The tables do not line up otherwise: no arithmetic relates the two numbers, and most parameters are named differently in each - LFO 1 Delay Time against LFO 1 Delay / Fade, VCA Envelope Attack against VCA Envelope Attack Time - so they are carried as separate controls, because pairing two names is a reader's judgement and not the maker's. Four names are printed in both tables and also twice in one of them, and none of those is paired.

FIVE NAMES ARE PRINTED TWICE, AND EACH IS CARRIED AS PRINTED. The controller table gives VCA Envelope Sustain to 40 and 44 and VCA Envelope Sustain Curve to 52 and 56; the NRPN table gives Attack Curve twice to each envelope, at 58 and 61, 67 and 70, and 76 and 79. Each second one sits where its neighbours put another parameter - 44 between the filter envelope's decay at 43 and release at 45, 61 where the VCA envelope's release curve would follow its sustain curve - and none is corrected: a definition says what the maker printed, and these say it twice. Their keys carry their numbers so that each can be told apart. The controller table's Replaces: remarks name the standard controller each number displaces, and from 43 to 52 they name the one a row below by the MIDI specification's list, which neither manual prints: 43 is given Replaces: Pan (LSB), which that list puts at 42. No field rests on a remark.

ONE RANGE IS READ FROM ITS OWN NOTES, AND TWO FROM A NEIGHBOUR. LFO 1 Mono Mode prints 0-1 in its Range cell and, in its Notes, "Poly (0) Mono (1) or SPREAD-1(2) to SPREAD-254(255)" (manual p. 136), which cannot both be true; LFO 2 Mono Mode prints 0-255 beside the same note, and the LFO pages describe the setting for both: "POLY - All voices use an independent LFO. This is the default mode." (manual p. 62), "MONO - All voices use a common LFO." (manual p. 62) and "The PHASE SPREAD range is from 1-254 degrees." (manual p. 62). That page prints no value range, and it describes the values the Notes number, so 0-255, the Notes' own numbering, is carried as a reading it supports. FX 2 Output Gain and FX 3 Output Gain print a dash for their range, between FX 1's and FX 4's 0-150, and FX 4's note reads "Output gain for FX slots 1-4" (manual p. 139); 0-150 is carried for both, as a reading of that note.

35 NRPNS CARRY THEIR NAMED VALUES, wherever the Notes cell names every value in the range and nothing else - an LFO's seven shapes, the thirteen polyphony modes, the eleven arpeggiator modes. Notes that name some values and not others, such as 20 bpm (0) to 275 bpm (255), are left as ranges. Twenty-four rows' Notes point to a list printed elsewhere, such as See Modulation Matrix Source List, and none of them carries names. Twenty-one of them carry no range either, the matrix's sixteen sources and destinations, the four FX Types and the Control Sequencer's Clock Divider: see below.

THE CHART IS ONE STAMP OVER TWO EDITIONS, AND THEY DISAGREE ABOUT TRANSPORT. Both print Version:1.0 and 01 September 2016 (manual p. 132; earlier_manual p. 113). The later fills in with Yes and No most of what the earlier leaves as dashes, and the two agree wherever both say something - except on Start, Continue and Stop: the earlier gives its Transmit cell as "Start: Yes, Continue: Yes, Stop: Yes" (earlier_manual p. 113) and its Recognize cell a dash, and the later gives No and Yes. Whether they are sent cannot be both, and no page outside the chart mentions them, so transport is not recorded. Song Position Pointer is No sent and Yes received in the later edition only.

WHAT ELSE THE CHART AND THE MANUAL'S PAGES GIVE: channels 1 to 16, 1-16 sent on the chart (manual p. 132), received on one channel or all - "The CHANNEL can be ALL, or any number from 1-16." (manual p. 26) - and sent on one; program change both ways, a setting whose fullest option is "RX-TX - Program changes will be transmitted and received." (manual p. 26); 1,024 programs, "1024 Programs arranged in 8 Banks of 128 Programs." (manual p. 6); note-on and note-off velocity both ways, the release velocity used only as a modulation source - "The NOTE OFF VELOCITY is only available within the MOD MATRIX as a MODULATION SOURCE." (manual p. 30); channel pressure and not polyphonic, both ways by the chart (manual p. 132), and sent by the keyboard: "Note: The DeepMind 12 keyboard sends 'Channel Aftertouch,' meaning the MIDI message created is common to all keys." (manual p. 30); pitch bend both ways, two semitones each way by default - "default P.BEND RANGE+ setting is +2 (notes)." (manual p. 78), and -2 down - and each direction settable from -24 to +24 on the POLY menu (manual p. 78), where NRPNs 36 and 37 print 0-24; and system exclusive both ways, with its own chapter from p. 140. MIDI clock is received from the port the arpeggiator's CLOCK names, MIDI or USB and MIDI(Auto) by default, falling back to the internal clock when none arrives, and sent when that is internal and TRANSMIT-CLK is on (manual p. 58).

CROSS-CHECKED AGAINST THE GUIDE BEHRINGER SERVES, a V 5.0 quick start guide, whose specifications agree with the manual's on everything a field here rests on: "Presets 8 banks each with 128 programs" (quick_start p. 43), "Fader Modes MIDI CC, NRPN, OFF" (quick_start p. 43) and "MIDI In/Out/Thru 5-pin DIN / 16 channels" (quick_start p. 43), and twelve voices, aftertouch, and note-on and note-off velocity. They differ in two counts: the guide gives the modulation matrix "22 sources / 130 destinations" (quick_start p. 43) and its effects 34 Algorithms, where the manual's specifications give "24 sources / 132 destinations" (manual p. 131) and 35 Algorithms. The earlier edition's own specifications page agrees with the guide, "22 sources / 130 destinations" (earlier_manual p. 112) and 34 Algorithms.

AND THE CITED EDITION'S OWN LISTS OUTGROW ITS NRPN TABLE. Its numbered lists run sources 1-24 and destinations 1-132 (manual pp. 93-94), where the earlier edition's run 1-22 and 1-129 (earlier_manual pp. 77-78). Both editions' NRPN tables give each matrix slot's source 0-22 and destination 0-129 (manual pp. 137-138, earlier_manual pp. 118-119) - the earlier lists' numbers and one value more, at 0. So the table was not raised with the lists, and as printed it stops short of the cited edition's: by its last two sources and last three destinations if a value is the list's own number, and by one and two if the values count from nought, which no page says. The source list is renumbered as well as lengthened - Expression is new at 6, where the earlier list has LFO1 - so a value names a different source in each edition, and no control here names them. The sixteen source and destination NRPNs say range: unknown. Their numbers are not in doubt, and their ranges are. The table's cannot reach the lists it points to, and whether the table or the lists describe firmware 1.1.2 no document says. Until 2026-10-12 they carried the table's ranges, because a control with no range reads as 0-127 and this format then had no way to say that a range is not established.

THE FOUR FX TYPES FALL SHORT THE SAME WAY, AND SAY range: unknown TOO. Each prints 0-33 with the note See FX Type list, in the cited edition (manual p. 139) and in the earlier one (earlier_manual pp. 119-120). The earlier edition's Effects Table runs 1-34 (earlier_manual p. 79), which thirty-four values would fit, and the cited edition's runs 1-35 (manual p. 95): Vintage Pitch is new at 34 and RotarySpkr moves from 34 to 35. So the table was not raised here either, and a value may name a different effect in each edition. Its thirty-four values cannot reach the thirty-fifth effect of the list it points to, so the four state no range, for the reason the matrix's sixteen do not. A slot can also hold no effect at all: the FX OVERVIEW screen shows each slot's TYPE as None (manual p. 22, earlier_manual p. 21), which neither Effects Table numbers and no page gives a value for. If None is one of the row's values, the table stops two short of its list and not one. Until 2026-10-13 the four carried [0, 33] as printed.

THE CONTROL SEQUENCER'S CLOCK DIVIDER FALLS SHORT BY FOUR, AND SAYS range: unknown. Its row prints 0-15 with the note See Control Sequencer divider table, in the cited edition (manual p. 138) and in the earlier one (earlier_manual p. 119). The table it points to lists twenty divisions in both, from 4 down to 1/64 (manual p. 55, earlier_manual p. 46), and the appendix's timing table has seventeen rows (manual p. 150, earlier_manual p. 133), so no divider or timing table in either edition has sixteen. Until 2026-10-14 it carried [0, 15] as printed.

FOUR MORE DISAGREEMENTS, NONE REACHING A FIELD. The Global Commands' Pedal CC lists "Foot Control (CC4), Breath (CC3), Volume (CC7), Expression (CC11)" (manual p. 139), where the controller table has Breath Controller (MSB) at 2 and a dash at 3. The chart gives every RPN No both ways (manual p. 132), where the controller table marks 100 and 101 Yes both ways. And the chart answers the modes with a dash and a pointer to the channel settings, where the table marks 124 to 127 Yes both ways. And the earlier edition's modulation list gives the 3D controller's axes as CC X (114) to CC Z (116) (earlier_manual p. 77), where both editions' tables put them at 115-117 and the cited edition's list agrees with its table. Controllers 2 and 115-117 are carried as the tables print them, and 100, 101 and 124-127 are not controls.

TWELVE VOICES, AND THE MODE THAT SHARES THEM IS A PROGRAM SETTING. "Number of voices 12 analog" (manual p. 131), one to a note in Poly: "Poly - A single voice is allocated to each note." (manual p. 75). The thirteen modes are NRPN 85. Poly-6 and Poly-8 give six and eight notes; the mono and unison modes give voices per note and not notes, "Unison-2 - The next available two voices are allocated to each note played." (manual p. 75), so the note count of each would be arithmetic on twelve or a reading of the word mono, and voicing_modes is not recorded.

WHAT IT SENDS, AND WHAT CHANGES IT. The arpeggiator's notes go out only with ARP-TO-MIDI on, "The default ARP-TO-MIDI setting is Off." (manual p. 58). An app that announces itself switches a port to NRPN: the control mode is "set automatically when an app notify message is received." (manual p. 139). The selected NRPN is kept for each port separately (manual p. 136). And a value above 127 needs both data entry bytes, since "The Data Entry (MSB) message is optional and can be ignored for switch parameters or any parameters which have a range of 0-127 or less." (manual p. 136); how a 0-255 value divides between the two bytes the page does not say.

NOT RECORDED, ALSO: mode, which the chart answers with "See RX/TX CHANNEL PARAMETERS for each MIDI interface" (manual p. 132); the note range, which the chart gives as C0-C8 sent and a dash received; the Global Commands, a table of global settings with ranges and no message number (manual p. 139); and groups, which neither table prints - the manual's edit menus group the panel, and placing 269 rows among them would be a reading on every row.