Detailed feature reference
Play the changes guide
CADENCE — PLAY THE CHANGES Chord lab → Follow the changes → Play in time. Start with Dm7 → G7 → Cmaj7: one bar, one bar, two bars in 4/4. Choose roots, thirds and sevenths, shells, or complete chords; 40–160 bpm; 1–4 passes. The default is two passes of guide tones at 80 bpm. Connect and select a USB MIDI keyboard, release the keys, then Count in & play. Keep going after a mistake. The current chord and next change stay visible. Common tones can remain held while the other voices move. Inversions and octave doublings are accepted. Shell and complete modes honor a printed slash bass. The bass accompaniment is optional. Turning off Keep the click still leaves the count-in. Audio comes from the Mac; it is not sent back as MIDI input. Hear an example plays a suggested voicing sequence without grading or saving. It also turns on tone hints. Turn hints off when you want to try independent recall; your result records whether hints and bass were enabled. RESULTS Harmony matched means the correct physical key set settled for at least 60 ms near the change. It does not require all notes to arrive in the same packet. On time means that settled shape first arrived within ±100 ms of the beat. An unchanged chord can stay held across a repeated symbol. N.C. asks for release at the entrance. The pedal does not substitute for held keys in this check. The result separates matched harmony, timing, missing/extra notes and recoveries. Extra presses near the change still count even if you also found the shape. Assessment windows are at most 600 ms each side of a change, reduced for closely spaced changes; they do not overlap. Playing well outside a window can appear as an unmatched chord. A shape already held from the count-in is early. Focus on this transition offers the difficult change with the preceding one, 10 bpm slower. It retains complete source bars, including any other changes inside those bars, to keep the click aligned with the original meter. Return to full progression restores the full selection. This suggestion is based on the first transition that did not arrive cleanly, not a diagnosis of technique. REVIEW AND SAVING Completed passes save a detailed result, journal summary and review schedule in one local operation. If saving fails, the result remains in memory: try saving again, or explicitly discard it. Leaving the view stops an unfinished pass; it does not count as a failure. Pending results survive navigation, not quitting. Learn → Practice & review → Return to the changes shows due reviews. A difficult pass returns after 10 minutes; a clean supported pass after 20 minutes. A clean independent pass starts at one day, with later successful, due reviews increasing spacing. Immediate retries cannot stretch the interval. Each score revision, selection, harmony staff, mode, tempo, loop count, focused passage and support/offset setup has its own review. A slower focused success does not erase the full-pass miss. These are simple scheduling rules, not a calibrated measure of mastery. YOUR OWN PIECES Use Follow changes from Compose, Library or Studies, then Play in time. The exercise uses the written chord symbols; it does not infer harmony from notes. The first version accepts 1–8 complete bars in steady 2/4, 3/4 or 4/4, at least half a second per change, and up to 90 seconds per pass session. Guide and shell modes require a third and seventh. Choose another mode for triads or sus chords. A preceding symbol carries into the selection. Before the first written symbol, the exercise treats the passage as N.C. Part and harmony staff selection remain attached to the result. Current piece results can inform coaching when you choose to include practice history; renaming retains evidence, musical edits invalidate its use for the new score. Practice copies resolve to their parent composition. TIMING LIMITS Use built-in or wired audio. Optional input offsets compensate for a consistent arrival delay; the eight-tap check in Practice can estimate a value for your setup, including your own tapping tendency. Offsets reset on device changes or reopening a view. Native MIDI source timestamps are preferred; arrival-time fallback is explicitly supported evidence, not independent timing recall. Keyboard changes, reset/panic, app focus loss, sleep and audio changes stop the pass without grading it. This checks entrances and physical key collections. It does not grade note length, sustained pedal sound, dynamics, technique, swing or expressive rubato, nor everything played between changes. Native virtual MIDI and scheduled audio checks pass; the physical Roland FP-10 remains to be tested.