Skip to content

Command Index

The complete command tree. For grouped, example-rich help run yazses --help or yazses <command> --help. See also the prose CLI reference.

yazses about

Show author, version, links, and where to report issues or request features.

yazses acronyms

Manage a persistent acronym glossary and expand acronyms in text — offline.

yazses acronyms add

Register an acronym → full-form expansion in your glossary.

  • Arguments: acronym, full

yazses acronyms expand

Expand known acronyms on first use ('Full Name (ACR)'), contract after — offline.

  • Arguments: text

yazses acronyms list

Show your stored acronym glossary.

yazses acronyms remove

Remove an acronym from your glossary.

  • Arguments: acronym

yazses audio

See and pin the input microphone (fixes a mic that silently switches).

yazses audio devices

List capture-capable audio input devices.

yazses audio status

Show the pinned mic, the OS default, and (if running) live capture health.

yazses audio use

Pin the input microphone by name so a monitor/headset can't steal capture.

  • Arguments: name
  • --clear — Unpin; follow the OS default.

yazses autostart

Start YazSes automatically at login, so it survives a reboot.

yazses autostart disable

Stop starting YazSes at login. The daemon keeps running until you stop it.

yazses autostart enable

Install and enable the login service, so YazSes is running when you sit down.

yazses autostart status

Say whether YazSes will be running after the next reboot.

yazses biosignal

EEG/EMG/EOG switches: muscle squeeze, double blink, jaw clench -- check, calibrate, watch, measure.

yazses biosignal boards

List the EEG/EMG boards the switches are known to work with and their BrainFlow ids.

  • --json — Print one JSON object instead of a table.

yazses biosignal calibrate

Guided calibration: learn YOUR resting level and your squeeze, blink and clench size.

  • --source — Calibrate on this source instead of [biosignal] source.
  • --blink, --no-blink — Include the blink step (default: [biosignal] double_blink).
  • --clench, --no-clench — Include the jaw-clench step (default: [biosignal] jaw_clench).
  • --yes, -y — Start without waiting for Enter.

yazses biosignal check-source

Check that a device or recording keeps the source contract the detectors rely on.

  • Arguments: source
  • --seconds, -s — How long to drive the source.
  • --json — Print one JSON object instead of a report.

yazses biosignal doctor

Check BrainFlow, serial ports, your calibration, and run a synthetic self-check.

  • --boards — List every BrainFlow board id, not just the relevant ones.
  • --source — Check this source instead of [biosignal] source.

yazses biosignal evaluate

Score the detectors against known events: precision, recall, latency, false alarms per hour.

  • --source — synthetic[:emg|eeg|eog|mixed]; ignored with --file.
  • --seconds — Length of the synthetic session.
  • --seed — Seed for the scripted gestures and noise (same seed, same result).
  • --file — Score this recording (CSV or .npz) instead; needs --truth.
  • --truth — JSON list of events you labelled: [{"kind": "double_blink", "t": 12.3}].
  • --tolerance — Seconds a detection may differ from the true time and still count.
  • --emg, --no-emg — Include the muscle-squeeze switch (default on; needs a calibration profile).
  • --blink, --no-blink — Include the double-blink switch (default on).
  • --clench, --no-clench — Include the jaw-clench switch (default on).
  • --json — Print one JSON object instead of a table.

yazses biosignal field-test

Guided test of how often the switches fire by mistake during ordinary activity.

  • --source — Test this source instead of [biosignal] source (synthetic | file: | brainflow:[:]).
  • --minutes — Stretch or shrink the test to about this long (0 = the standard length).
  • --json — Print one JSON object instead of text.
  • --out, -o — Also write the counts to this JSON file (owner-only; never overwrites).

yazses biosignal monitor

Watch signal quality and detections live. Nothing is typed or acted on.

  • --source — Watch this source instead of [biosignal] source (synthetic | file: | brainflow:[:]).
  • --seconds — Stop after this long (0 = until Ctrl-C).
  • --interval — Seconds between status lines.
  • --emg, --no-emg — Override [biosignal] emg for this run.
  • --blink, --no-blink — Override [biosignal] double_blink for this run.
  • --clench, --no-clench — Override [biosignal] jaw_clench for this run.

yazses biosignal profile

Show the saved calibration profile: board, channels, age. Never an amplitude.

  • --json — Print one JSON object instead of text.

yazses biosignal profile-clear

Delete the saved calibration profile.

  • --yes, -y — Delete without asking.

yazses biosignal record

Record raw signal to a local file, to replay or evaluate later.

  • --out, -o — File to write: .csv (streamed) or .npz.
  • --source — Record this source instead of [biosignal] source (synthetic | file: | brainflow:[:]).
  • --seconds — How much signal to record.
  • --force — Overwrite --out (and --truth-out) if it exists.
  • --seed — Synthetic source only: seed for the scripted gestures and noise.
  • --truth-out — Synthetic source only: also write the scripted gestures as a truth file for evaluate.
  • --json — Print one JSON object instead of text.

yazses biosignal replay

Run a recording through the pipeline and print each detection and the false-positive summary.

  • Arguments: file
  • --emg, --no-emg — Override [biosignal] emg for this run.
  • --blink, --no-blink — Override [biosignal] double_blink for this run.
  • --clench, --no-clench — Override [biosignal] jaw_clench for this run.
  • --json — Print one JSON object instead of a table.

yazses biosignal report

Show your own measured activation and cancel rate from the local counters.

  • --json — Print one JSON object instead of a table.

yazses braille

Translate text to Unicode Braille (UEB subset) — fully offline.

  • Arguments: text
  • --grade — UEB grade: 1 (uncontracted) or 2 (contracted).

yazses case

Recase text to a naming convention — fully offline.

  • Arguments: text
  • --style — Target case: snake|kebab|camel|pascal|title|sentence|upper|lower|constant. Omit to detect a spoken 'make this … case' command in the text.

yazses chords

Turn a spoken key chord into injectable key combos — fully offline.

  • Arguments: text

yazses cite

Resolve a spoken 'author year' reference against a .bib file — fully offline.

  • Arguments: query
  • --bib — Path to a BibTeX (.bib) file.
  • --style — latex | plain | apa.

yazses cliphistory

A persistent clipboard history you can recall by voice-style reference — offline.

yazses cliphistory add

Add an entry to your clipboard history (newest first, de-duplicated, capped).

  • Arguments: text

yazses cliphistory list

Show your clipboard history, newest first.

yazses cliphistory recall

Print the history entry a spoken reference points to — offline.

  • Arguments: query

yazses coach

Show private speaking-style analytics (filler rate, words-per-minute, vocabulary).

  • --limit, -n — How many recent dictations to analyse.

yazses corpus

Inspect or clear the local learning corpus.

yazses corpus destroy

Irreversibly delete the corpus database (corpus.db) and its audio clips (clips/).

  • --i-mean-it — Required: confirm the irreversible delete.

yazses corpus forget

Delete recently captured events (e.g. after dictating something private).

  • --minutes, -m — Delete events from the last N minutes.

yazses corpus status

Show the learning corpus size, event counts, and date range.

yazses doctor

Check system prerequisites and report what's OK / missing.

  • --mic — Also record a short ambient clip and compare its level to the VAD threshold.

yazses enroll

Run the accessibility enrollment wizard to calibrate VAD thresholds.

yazses enroll-voice

Record a short voiceprint so YazSes can recognise your voice.

yazses eye-eval

Run an eye-control evaluation task and save a privacy-safe result you can inspect.

  • Arguments: task
  • --out, -o — Where to write the result JSON.
  • --synthetic — No-camera dry run: generate the outcomes instead of measuring them.
  • --outcomes — A JSON list of per-trial outcomes you recorded during a real run.
  • --blocked — Record a BLOCKED result. One of: not_supported, not_measured, permission_denied, tracking_unavailable, participant_stopped, technical_invalidation.
  • --study-mode — ci | synthetic | community_qa | research. Never inferred.
  • --protocol-id — The named, versioned research protocol. Required by --study-mode research.
  • --camera-class — integrated | external | virtual | none. Required with --outcomes.
  • --perception-backend — Which camera perception backend ran, e.g. mediapipe.
  • --verdict — Your own PASS/PARTIAL/FAIL/BLOCKED, overriding the computed one.
  • --seed — Task seed. Default: the seed the task fixture records.

yazses facegesture

Face-gesture switch — capture your relaxed face so the gesture is judged against it.

yazses facegesture calibrate

Capture your neutral (relaxed) face as the baseline for a face gesture.

  • --gesture — Gesture to calibrate (default: [facegesture] gesture).
  • --seconds — How long to watch your relaxed face.
  • --yes, -y — Start without waiting for Enter.
  • --active — Also capture your ACTIVE level: a few deliberate holds of the gesture. Needed when your resting level is near zero and the default thresholds never fire.
  • --talk — Also capture your TALK level: speak normally so the derived hold threshold stays above it. Needs an active level (from this run or stored).
  • --skip-rest — Reuse the stored resting baseline; capture only --active and/or --talk.
  • --holds — Deliberate holds for --active.
  • --hold-seconds — Seconds each --active hold is measured.
  • --talk-seconds — Seconds of speech for --talk.

yazses facegesture reset

Forget the stored neutral baseline(s) and active/talk levels; back to configured thresholds.

  • --gesture — Forget only this gesture's baseline.

yazses facegesture status

Show what the face switch's select, confirm and cancel gestures last did.

yazses features

See capabilities and turn them on/off (no config-editing needed).

  • --on — Show only capabilities that are currently ON.
  • --tier — Filter by advice tier: core, on, rec, opt, exp.
  • --category, -c — Filter by category name (partial, case-insensitive), e.g. 'access'.

yazses features disable

Turn a capability OFF (writes your config), then yazses restart to apply.

  • Arguments: name

yazses features enable

Turn a capability ON (writes your config), then yazses restart to apply.

  • Arguments: name
  • --force — Allow enabling experimental features.
  • --no-install — Don't auto-install the feature's optional deps.
  • --yes, -y — Don't ask before downloading an add-on (Windows installer build only).

yazses features info

Describe a capability — what it does, a usage example, and how to toggle it.

  • Arguments: name

yazses features install

Install the libraries that switched-on features are missing.

  • Arguments: names
  • --yes, -y — Install without asking first.

yazses features repair

Repair every switched-on feature whose libraries are missing (shown as ON, not installed).

  • --yes, -y — Install without asking first.
  • --dry-run — List the dormant features and their packages; install nothing.

yazses features reset

Restore every capability to the state a fresh install ships with.

  • --yes, -y — Skip the confirmation prompt.
  • --dry-run — List what would change and write nothing.
  • --no-install — Don't auto-install optional deps for what it turns on.

yazses fileopen

Open a file by voice: fuzzy matches your spoken query against files in a directory.

  • Arguments: query
  • --dir, -d — Directory to search in.
  • --yes, -y — Launch immediately without confirmation.

yazses findreplace

Apply a spoken find-and-replace to text — fully offline.

  • Arguments: command
  • --in — Text to apply it to (omit to read stdin).

yazses gaze

Aim dictation with your gaze — type into whichever pane you look at.

yazses gaze calibrate

Calibrate the webcam so your gaze maps to screen zones.

  • --no-install — Don't auto-install the gaze deps (mediapipe, opencv).

yazses gaze measure

Measure how well your saved gaze calibration works, and write a report to share.

  • --camera-class — integrated | external | virtual. Never guessed.
  • --out, -o — Where to write the report JSON (default: ./yazses-gaze-report.json).
  • --rounds — How many validation passes to run.
  • --interval-s — Seconds to wait between passes (to see drift over time).
  • --per-point — Gaze samples averaged per target.
  • --block-s — Drift bucket length in seconds.
  • --study-mode — community_qa | research. Never inferred.
  • --protocol-id — The named, versioned protocol. Required by --study-mode research.
  • --camera-label — A short model name of your choosing, e.g. 'Integrated 720p'. No serial numbers.
  • --glasses — none | glasses | contacts | not_stated.
  • --lighting — dim | normal | bright | backlit | not_stated.
  • --viewing-distance-cm — Eyes to screen, in centimetres.
  • --screen-diagonal-in — Screen diagonal in inches. With --viewing-distance-cm, errors are also given in degrees.
  • --routing-attempts — Dictations you tried to aim with gaze (your own count).
  • --routing-fallbacks — Of those, how many fell back to the focused window.
  • --wrong-target — Of those, how many you had to correct (landed in the wrong window).

yazses gaze rollback

Undo the last implicit refinement and restore the calibration it replaced.

yazses gaze status

Show whether look-to-pane is ready: deps, desktop backend, calibration.

yazses gesture

Control Gesture Chords: see what is bound and why nothing fires, pause or resume them.

yazses gesture pause

Stop chords from acting. The pause chord itself still works.

yazses gesture resume

Resume after a pause, a fault or a stale camera. Chords never fire from before it.

yazses gesture status

Show whether Gesture Chords are running, and if not, why not.

yazses gitvoice

Turn a spoken git command into a git command — fully offline.

  • Arguments: text
  • --run — Run the command instead of only printing it.
  • --yes — Confirm a destructive command so --run will run it.

yazses handsfree

Turn on the hands-free preset; stop, re-arm and inspect every camera-driven input at once.

yazses handsfree disable

Turn the hands-free preset OFF: stops the pointer and switch now, frees the camera on restart.

yazses handsfree enable

Turn the hands-free preset ON: one switch that sets the ordinary feature configs.

  • --force — Allow the experimental camera features the preset turns on.
  • --no-install — Don't auto-install the camera features' optional deps.

yazses handsfree explain

Print the ordinary feature settings the hands-free preset sets, without applying anything.

yazses handsfree pause

Stop all hands-free input: no pointer motion, no click, no face-switch key press.

yazses handsfree resume

Re-arm hands-free input after a pause or a fault. The first frame afterwards never acts.

yazses handsfree status

Show the same panel the settings window shows: state, health and what to do next.

yazses headpointer

Control the Head-Pointer: recenter, pause or resume it, hold a clutch, toggle dwell-click.

yazses headpointer clutch

Hold the pointer still without turning the feature off (e.g. while you pull a face).

  • --off — Release the clutch instead of holding it.

yazses headpointer dwell

Turn dwell-click on or off without touching pointer motion.

  • Arguments: state

yazses headpointer pause

Stop the pointer: no motion, no dwell progress, no click.

yazses headpointer recenter

Make your current head position the neutral one. No motion and no click while it happens.

yazses headpointer resume

Resume after a pause or a fault. The first frame afterwards never moves or clicks.

yazses headpointer status

Show whether the Head-Pointer is running, and if not, why not.

yazses hotkey

Change the key you hold to talk.

yazses hotkey command

Set a dedicated command key, then yazses restart to apply.

  • Arguments: key

yazses hotkey set

Set the key you hold to dictate, then yazses restart to apply.

  • Arguments: key

yazses hotkey show

Show the current hold-to-talk key (and command key, if set).

yazses inject

Type text into the focused window without recording (tests the injector).

  • Arguments: text
  • --delay, -d — Seconds to wait before injecting, so you can focus the app to test.
  • --diagnose — Report the focused window before injecting (Windows only).

yazses inject-backend

Show or pin the keystroke-injection backend.

  • Arguments: name
  • --allow-portal — Let auto use the desktop portal on Wayland.
  • --deny-portal — Never use the desktop portal.

yazses jump

Jump to a symbol or line in the active editor.

  • Arguments: target

yazses language

Inspect dictation-language profiles and effective STT language state.

yazses language list

List the high-level language profiles YazSes can resolve coherently.

yazses language set

Switch dictation language through the shared transactional service.

  • Arguments: profile
  • --dry-run — Show the complete plan and prerequisites without changing anything.
  • --model — Explicit faster-whisper model override.
  • --engine — Explicit engine override (currently only faster-whisper is accepted).
  • --recommended-model — Select the profile's supported baseline instead of preserving a compatible model.
  • --no-download — Never fetch a missing speech model; refuse before writing config.
  • --no-install — Never install a missing optional Python dependency.
  • --no-restart — Commit a valid config but leave a running daemon unchanged.
  • --yes, -y — Apply the displayed plan without an interactive confirmation.

yazses language status

Show the speech language, Han script, engine/model, multilingual support, and coherence.

  • --json — Emit a machine-readable status object.

yazses logs

Show the daemon's diagnostic log (metadata only unless log_level is DEBUG).

  • --lines, -n — Number of recent lines to show.
  • --path — Print the log file path and exit.

yazses mark-wrong

Flag the last dictation as a misrecognition (a learning signal).

  • --correction, -c — What you actually said (optional).

yazses mcp-server

Expose YazSes to another agent over MCP, on stdin/stdout (ADR-020).

yazses meeting

Hands-free meeting recording with who-said-what speaker labels + notes.

yazses meeting enroll

Enroll a meeting speaker as a named voiceprint so they're auto-named next time.

  • Arguments: meeting_id
  • --speaker — Cluster id to enroll, e.g. speaker_1.
  • --name — The person's name, e.g. Alice.

yazses meeting list

List stored meetings on this machine (no daemon required).

  • --json — Print meetings as a JSON array (machine-readable; empty list is []).

yazses meeting notes

Generate meeting minutes (summary, decisions, action items) from the transcript.

  • Arguments: meeting_id
  • --force — Write notes even for a recording that holds no speech.

yazses meeting recover

Re-run the post-pass on a meeting whose finalize never completed.

  • Arguments: meeting_id
  • --force — Re-run even a meeting that finished cleanly (its outputs are archived, never overwritten).

yazses meeting relabel

Fix speaker labels — merge two clusters and/or rename them — and re-render.

  • Arguments: meeting_id
  • --merge — Fold one speaker into another: SPEAKER_2=speaker_1 (repeatable).
  • --rename — Name a speaker: speaker_1=Alice (repeatable).
  • --format, -f — Transcript format to re-render.

yazses meeting start

Start recording a meeting (hands-free — no key to hold).

  • --speakers, -s — How many people are in the room. On this backend it is an exact count, not a maximum, so leave it at 0 unless you know the number. 0 = use [meeting] max_speakers.

yazses meeting status

Show the running meeting (elapsed + live transcript) or recent meetings.

yazses meeting stop

Stop the recording and generate the speaker-labelled transcript (and notes).

yazses meeting summary

Show what a meeting produced, where each file is, and what not to trust.

  • Arguments: meeting_id

yazses mic-level

Measure your room and your voice, and put the VAD threshold between them.

  • --seconds, -s — Seconds to record while you SPEAK (the room is measured for half this).
  • --set — Write the recommended vad_threshold to config.

yazses model

Manage SLM intent-routing models (download / list).

yazses model download

Download a model: a speech-to-text checkpoint, or a GGUF for intent routing.

  • Arguments: model_id

yazses model list

List available models (speech-to-text and SLM) and their status.

yazses outline

Build a nested outline incrementally and render it to Markdown/OPML — offline.

yazses outline add

Add an item after the cursor (same level) and show the outline.

  • Arguments: text

yazses outline clear

Discard the current outline and start fresh.

yazses outline indent

Indent the current item one level deeper (max one deeper than the item above).

yazses outline promote

Promote the current item one level shallower (floor 0).

yazses outline render

Render the current outline to Markdown or OPML.

  • --format, -f — markdown | opml.

yazses overlay

Run the sonar voice-activity overlay (needs the overlay extra: PySide6).

yazses packs

Add-ons: a feature's libraries, downloaded and verified (Windows installer build).

yazses packs clean

Delete set-aside, half-finished and old builds' add-ons (keeps what rollback needs).

yazses packs install

Download, verify and install one add-on for this build. Writes no config.

  • Arguments: name
  • --yes, -y — Don't ask before downloading.
  • --from — Install this .yazpack file instead of downloading (offline). Its name and SHA-256 must match this build's built-in add-on list.

yazses packs path

Print the folder add-ons are installed in.

yazses packs probe

Mount one unpacked add-on and import its entry modules (used by install and CI).

  • Arguments: name
  • --staged — The unpacked add-on folder to test.

yazses packs remove

Delete an installed add-on. Features that need it then read waiting for add-on.

  • Arguments: name
  • --all — Remove every add-on of every build.

yazses packs verify

Check every installed file's size and SHA-256. A damaged add-on is set aside.

  • Arguments: name

yazses punch-in

Correct the last dictation by re-speaking just the wrong phrase.

  • --dry-run — List candidate spans without editing (confirm first).
  • --choose, -n — Apply the candidate at this rank (0 = best).

yazses quickstart

Get dictating in 3 steps — a friendly guide tailored to your machine.

yazses recall

Search your past dictations.

  • Arguments: query

yazses reflow

Reflow a monologue into a bulleted outline — fully offline.

  • Arguments: text

yazses remote

Forward voice typing to a remote host over SSH.

  • Arguments: host
  • --port, -p — SSH port.
  • --key-file, -i — Path to SSH private key.
  • --stop — Disconnect active remote session.

yazses report

Collect a diagnostic report locally — nothing is uploaded.

  • --output, -o — Where to write it.
  • --print — Print it instead of writing a file.
  • --log-lines — How much log tail to include.

yazses restart

Restart the daemon — kills any stray/duplicate daemons and starts exactly one.

  • --force — Restart even while a meeting is still being written up.

yazses say

Speak text aloud with the built-in offline voice.

  • Arguments: text

yazses scratch

Show or clear your ambient scratch notes (spoken "note to self …").

  • Arguments: action

yazses screenplay

Format dictated lines as Fountain screenplay markup — fully offline.

  • Arguments: text

yazses send-to

Open a desktop AI app with your text filled in — never sent; you press Enter.

  • Arguments: client, text

yazses settings

Open the Settings window — every capability as a toggle, grouped by category.

  • --install-launcher — Add 'YazSes Settings' to your desktop app grid, then exit.
  • --uninstall-launcher — Remove the app-grid entry and its icons, then exit.

yazses setup

Provision all Linux runtime requirements so dictation works out of the box.

  • --dry-run — Show what would be installed/changed without doing it.

yazses shellpipe

Render a spoken pipeline into a shell command — fully offline.

  • Arguments: text

yazses slotfill

Extract structured fields from one utterance by a schema — fully offline.

  • Arguments: text
  • --slot — A field spec (repeatable): NAME:after=kw1,kw2 or NAME:choices=a,b,c.

yazses srs

Capture facts as spaced-repetition flashcards and schedule reviews (SM-2) — offline.

yazses srs capture

Capture a 'remember that X is Y' fact as a cloze flashcard — offline.

  • Arguments: text

yazses srs list

Show your flashcard deck and each card's next-review interval (days).

yazses srs review

Grade a card's recall (SM-2) and update its schedule.

  • Arguments: card
  • --grade, -g — Recall quality 0–5 (<3 lapses).

yazses staged

Speak, review, then commit — dictation lands in a buffer instead of typing.

yazses staged commit

Type everything pending into the focused app and clear the buffer.

yazses staged discard

Drop everything pending without typing it.

yazses staged status

Show what is pending review, and whether staged mode is on.

yazses staged undo

Remove the most recent staged burst (the same as saying "scratch that").

yazses start

Start the YazSes daemon (restarts cleanly if one is already running).

  • --no-autostart — Don't also set YazSes to start at login.

yazses status

Show daemon status. Queries the daemon over IPC when reachable.

  • --json — Output status as JSON for scripts and status bars.

yazses stop

Stop YazSes: the daemon, and the tray icon with it.

  • --force — Stop even while a meeting is still being written up.

yazses table

Turn spoken rows into delimited (CSV) lines — fully offline.

  • Arguments: text
  • --sep — Field separator for the output (default: comma).

yazses test

End-to-end self-test: confirm the injector works without speaking.

yazses transcribe

Transcribe an audio file to text — fully offline, on your machine.

  • Arguments: audio_file
  • --format, -f — Output format: txt (default) | md | srt | vtt | json (srt/vtt add timestamps).
  • --diarize, --no-diarize — Tag who said what with local speaker models (needs the diarization extra).
  • --speakers — Force an exact speaker count (0 = auto-detect).
  • --min-speakers — Ignored by the default sherpa diarizer — only the pyannote backend reads a lower bound (optional diarization-pyannote extra). The run warns and continues. Use --speakers to constrain the count.
  • --max-speakers — Force exactly this many speakers on the shipped diarizer (same as --speakers). 0 = auto-detect.
  • --names — Comma list mapped to speakers in order of first appearance: 'Alice,Bob,Carol'.
  • --rename — Explicit speaker→name map, repeatable: --rename speaker_0=Alice --rename speaker_1=Bob.
  • --language — Spoken language as a Whisper code -- 'en' (default), 'fa', 'de'; '' auto-detects; 'translate' renders any-language audio into English text. A non-English code needs a multilingual model (drop the .en suffix).
  • --model — STT model override, e.g. base.en (fast) or small.en (more accurate). Default: your config model.
  • --out, -o — Output file path (default: sidecar .).
  • --download-models — Download the ~45 MB sherpa diarization models, then exit (no transcription).

yazses tray

Show the system-tray / top-bar icon with a click-menu (mic + daemon controls).

  • --background, -b — Launch detached instead of blocking.

yazses tune

Analyze the learning corpus and propose accuracy improvements.

  • --apply — Review and apply proposals interactively.
  • --retranscribe, --no-retranscribe — Re-transcribe captured audio with a larger model to find errors.
  • --limit — Re-transcribe only the N most recent clips (0 = all). Budget roughly 10 s per clip on a laptop CPU — measured at 9.6 s with small.en — and a corpus at the default 500 MB cap holds ~1500, so a full run is several hours, not one. This is the middle ground between that and --no-retranscribe.

yazses update

Check for a newer YazSes and update it (snap / uv / pipx / pip / Windows).

  • --check — Only report whether an update is available; don't install.
  • --yes, -y — Install the update without prompting.

yazses verify

Record, transcribe, and prove dictation works end to end on this machine.

  • --seconds, -s — How long to record.
  • --type — Also type the transcript — into a small window YazSes opens, and check it arrived (into the focused window when no such window can open).

yazses vocab

Manage your personal dictionary (words STT mis-hears).

yazses vocab add

Add words to the dictionary so YazSes spells them right — no restart needed.

  • Arguments: words

yazses vocab export

Print your personal dictionary, one entry per line.

  • --output, -o — Write here instead of stdout.

yazses vocab import

Add entries from a file (or stdin) to your personal dictionary.

  • Arguments: source
  • --merge, --replace — Merge into the existing dictionary (default), or replace it entirely.
  • --yes, -y — Skip the confirmation for --replace.

yazses vocab list

Show the words in your personal dictionary.

yazses vocab remove

Remove a word from your personal dictionary — no restart needed.

  • Arguments: word

yazses wordfind

Reverse dictionary: describe a word and get candidates — fully offline.

  • Arguments: description
  • --limit, -n — Max candidates to show.
  • --lexicon — Path to a JSON {word: definition} lexicon to merge in.

yazses wordgoal

Track words written against a writing goal, across invocations — offline.

yazses wordgoal add

Add a chunk of text to your running word count and show progress.

  • Arguments: text

yazses wordgoal goal

Set (or clear, with 0) your writing goal.

  • Arguments: words

yazses wordgoal reset

Reset the word count to zero (keeps the goal).

yazses wordgoal status

Show your current word count and goal progress.