One Tensor preset variation sets the ON & HOLD switches to LATCH, and uses the HOLD – REC and DIR – FWD 3-way toggles. The idea is to tap HOLD to record the reverse audio into the buffer, then have it play back automatically when that buffer is full.
The REPEAT function in the TimeFactor / H9 will almost get you there (there’s a 5.8 second maximum buffer in the Tensor). But it’s a feature available within every TimeFactor delay preset. No need for finger toggling or external MIDI.
100% WET MIX will be normalized to always-on, rewinding phrases. Map the Aux switch (or a 2nd Aux) to the Performance / Repeat function. Here, REPEAT ON captures the last 2400 mS in the record buffer, and plays it back (reversed) in a continuous loop. REPEAT OFF drops that loop, and slides back into live input being rewound. To get to the 5.8 second buffer size in the Tensor, you’d have to turn to the Looper algorithm, or possibly FILTER PONG.