Your stats, your voice, your rules.
Your week
Finish a session to draw your first signal.
Session response
Answer out loud or unlock large touch choices. Set up practice for how you carry your phone.
Swipe to unlock choices
Hide touch answers until you swipe. Relocks for each question and after a pause. Voice stays available.
Speak every prompt
Read problems aloud while keeping voice answers available.
Weekly goal
How many runs count as a full week, for the Home progress bar.
Current speech path
Answer recognition
No model download needed. Your browser may use an online recognition service.
Question voice
Kitten Nano · compact English voice
Local neural speech with eight voices. Reads long cards in small chunks without using browser speech. English only; Kokoro remains available for Japanese.
Checking storage…
15 million parameters · INT8 · 24.4 MB weights plus runtime and voices. Download size is not peak RAM. Voice quality differs from Kokoro; try a preview on your phone.
Downloaded Kokoro files
Manage storage independently of the selected voice. Only one precision is loaded for playback. Download size is not memory use.
Kokoro Q8 · English + Japanese base pack
Checking storage…
Includes Q8 weights, voices, and pronunciation resources. Japanese pronunciation loads only when used. Removing this pack also removes all Kokoro precisions.
Q8F16
Checking storage… · 86.0 MB
Download the base pack first.
Q4
Checking storage… · 305.2 MB
Download the base pack first.
Q4F16
Checking storage… · 154.6 MB
Download the base pack first.
Experimental on iPhone. F16 variants are not guaranteed faster or smaller in memory. German uses system speech. Licenses and source information
Recovery & automatic backup
Your device stays primary. Encrypted cloud snapshots protect progress, settings, Anki cards, audio, and review schedules. Keep the latest 10 versions, up to 32 MB each.
Checking device storage…
Changes back up while the app is open and online. Queued uploads retry on reopening; iOS may stop background work. The last completed upload wins—other devices are not automatically overwritten.
Restore with a recovery phrase
Encrypted beta diagnostics
Enabled by default for session logs and practice microphone audio. Includes questions, transcripts, answers, cues, timings, device settings, and environment; never credentials. Encrypted on this device with keys derived from your recovery phrase. Recording and uploads require automatic backup to be enabled.
Reports are available for 14 days. Export decrypts the most recent 100 batches on this device into a readable troubleshooting file—share it only with someone you trust.
Audio clips capture what Moonshine or Whisper hears during practice, including speaker echo and nearby sound, and are available for seven days. Browser recognition does not expose its microphone audio. Recording stops when practice is paused or closed; no extra microphone is opened. Turning diagnostics off stops recording and discards pending logs and clips.
The device removes each encrypted clip after confirmed upload. Offline clips retry while open; the oldest are dropped at seven days, 256 clips, or the 64 MB device limit. Cloud audio is capped at 2 GB per recovery phrase.
No audio clips waiting to upload
Local data & file backup
Practice works from this browser’s data, including when cloud backup is off. Diagnostic audio follows the main diagnostics setting, is encrypted, and stays out of practice backups.
Covers preferences, session history, Planning skill and assessment schedule, and your arithmetic skill-ladder progress -- not imported expedition packs (re-import those from their original .apkg file).