Project Mad AdderHelp Centre
All help articles
Animals & careCurrent Beta guidance

Using PMA Capture with voice and quick entry

PMA Capture turns natural spoken or typed husbandry notes into reviewable structured actions so routine record entry can stay fast without silently writing guessed data.

Voice interpretation currently supports English only. PMA's voice vocabulary and language support can expand over time, but other languages are not claimed as supported today.

About 12 minutesessential

Before you begin

  • Use a current browser with microphone permission if you want voice capture. Typed Capture remains available when speech recognition is unavailable.
  • Open Capture from the most useful animal or Housing context when possible so PMA has a clear target.
  • Speak naturally. The examples below are patterns PMA understands, not exact “magic phrases” you must memorize.
  • Always review the detected action, animal, values, units, and outcome before confirming.

Interface walkthrough

These annotated interface maps describe the current PMA controls. They do not replace a real record review in your workspace.

1Voice or typed transcript
Look for: Record / Stop recording / transcript field

Speak continuously or type the care note yourself. Short pauses do not intentionally end an active recording session.

2Interpret actions
Look for: Interpret actions

PMA separates supported husbandry actions and prepares one or more reviewable drafts.

3Review and confidence
Look for: Detected action, fields, uncertainties

Correct anything PMA could not determine confidently. Ambiguous operational language stays reviewable instead of being silently treated as fact.

4Confirm and save
Look for: Confirm and save

Live records are written only after your explicit confirmation. Multi-action Capture keeps each proposed record separately reviewable.

Step-by-step workflow

1

Start in the right context

Open PMA Capture from the animal, Housing area, Smart Scan result, or global Capture control. Context helps PMA identify which record should receive the action.

2

Speak or type what happened

Describe completed care in normal English. You can include more than one action in a sentence—for example, feeding, then weight, then water—and PMA can split supported actions into separate drafts.

3

Use supported voice language

These are representative phrases currently recognized by the English interpreter:

Feeding“Fed him a live fuzzy rat and he ate it.” / “Offered a frozen-thawed hopper mouse and she refused it.”
Weight“Weighed her at 842 grams.” / “Weight was 1.2 kilograms.”
Length“Measured 36 inches long.” / “Length was 92 centimetres.”
Shed“Complete shed today.” / “He is in blue.” / “Eyes cloudy.”
Cleaning“Spot cleaned the enclosure.” / “Deep cleaned and changed substrate.”
Water“Changed their water.” / “Refilled the water bowl.” / “Topped up the water.”
Observation“Noticed she is more active today.” / “Observation: hiding more than usual.”
Environment“Hot side temperature was 88 degrees.” / “Humidity was 62 percent.”
Breeding observation“Pair locked.” / “Courtship observed.” / “Pair introduced.”
Inventory usage“Used 12 from inventory.” / “Took 6 from stock.”
Expense evidence“Spent 48 dollars on feeders.” / “Receipt for 32 dollars.”
Photo evidence“Photo.” / “Took a picture.” / “Add image evidence.”
Quarantine release“Released from quarantine.” / “Ended isolation.” / “Cleared containment.”

PMA also understands common wording variants around these actions. A general note can be retained when text is non-operational, while uncertain operational language is flagged for review.

4

Review every proposed record

Check the animal or Housing target, action type, quantities, measurement units, feeder details, outcomes, notes, and any uncertainty messages. Correct the draft rather than relying on the transcript alone.

5

Confirm only accurate records

Select Confirm and save only after the structured draft matches what actually occurred. Quarantine release and other containment-sensitive work keeps additional confirmation safeguards.

Expected result

One spoken or typed note becomes one or more explicit, reviewable PMA records, with uncertain fields visible before any live husbandry record is created.

Safety boundaries

  • Voice recognition is currently English-only and depends on browser speech-recognition support.
  • PMA Capture is an entry assistant, not an autonomous care system. It does not replace keeper review or professional veterinary judgment.
  • Negated or ambiguous operational language is deliberately handled cautiously so PMA does not record work that was not completed.
  • Quarantine release requires containment-state review and the appropriate permission.
  • Never confirm a measurement, feeding result, medication-related note, or other care evidence that the structured draft interpreted incorrectly.

Troubleshooting

Voice capture is unavailable

Use typed Capture, or try a current Chromium-based browser with microphone permission. Browser and device speech-recognition support varies.

PMA picked the wrong action

Edit or rewrite the transcript with clearer completed-action wording, then interpret again. Do not confirm the wrong action just to move forward.

A feeder phrase was transcribed strangely

PMA repairs several common English speech-recognition substitutions in feeder context, but you should still verify feeder type, size, preparation, quantity, and outcome.

One sentence created several drafts

That is expected when PMA detects multiple supported actions. Review each draft independently; failed actions can be retried without repeating successful ones.

My language is not recognized

Voice interpretation currently supports English only. Type the record in supported English for now rather than relying on an unsupported language.

Related articles