KontexVoX is simple to use, but a few good habits make the difference between a smooth session and a frustrating one. These ten tips come from the field — my own sessions, mistakes included.
Environment and recording
1. Lighting: in Capture Mode, make sure you have enough light for the photo to be sharp. The description will be readable, but the photo needs to be too.
2. Microphone distance: speak at normal volume, 20-30 cm (8-12 inches) from the phone. Too far, ambient noise wins; too close, it saturates.
3. Background noise: avoid very noisy environments (engines, loud music). The transcription is robust, not magic.
4. Inline corrections: if a word is mistranscribed, swipe left to go back and fix it right away. An error you let slide is an error you'll meet again in search.
Organizing the session
5. Sessions by batch: in Classic Mode, group your photos by theme before starting. One session = one theme = one coherent story.
6. Speak, then swipe: the slideshow moves when you swipe. Say what you have to say, then move on — don't let silence settle in.
7. No more than 50 photos per session: beyond that, vocal fatigue sets in and descriptions lose precision. That's exactly the threshold where the "by hand, on a keyboard" version would have made you quit at photo twenty — by voice, fifty goes fine, but not a hundred in a row.
8. Prepare your story: before you start, roughly know what you want to say for each group of photos. Improvising works, but the best descriptions come from structured thinking.
After the session
9. Export immediately: don't let enrichments sleep inside the app. AirDrop or Files, as soon as the session ends.
10. Check one photo at random: open an exported photo and confirm the description is there. Five seconds that prevent bad surprises — the full method is in Check that your photos are properly enriched.
Most important of all: speak naturally. No need to dictate like a robot — KontexVoX is designed for spoken voice, not dictated voice. Say "the torque wrench Dad gave me, top drawer" the way you'd say it to someone.
Frequently asked questions
What should a description say to be useful?
Think about future search: the words you'll type in three months. An object = its name + one distinctive detail + its location. A memory = who, where, and the detail nobody else knows.
The transcription gets some words wrong?
Proper nouns and technical vocabulary are the hardest for speech recognition. Fix them the first time they appear (tip 4) — and articulate slightly more on those words.
Can I pause in the middle of a session?
Yes — the session resumes where you left it. It's even recommended on big batches: twice twenty-five photos beat one tired fifty.