Dish in frame Weight · baseline
Is the dish itself well in view? You can only learn what you can see, so this is the baseline.
See It Work
Not a vague promise — the master's capture and the learner's training already run.
01 A System That Actually Runs
This page is not a how-to manual. It is the evidence that one connected flow — the master films, the AI learns the taste, and a learner in the U.S. masters the dish — is built to actually run, not just imagined. Below, see what each party does from their seat, and how the process is scored and learned, exactly as it is implemented.
The whole flow — from filming to on-site learning
Master · Korea The master films Films the standard cooking process from a Galaxy XR first-person view. Cloud AI learns the taste A high-intelligence general AI learns the dish's taste from the footage.
Learner · U.S. The learner trains live Seeing the learner's live field of view, the AI teaches by conversation and reports the deviation from the master as a grade. The next three sections show, in order: how the master films and is scored, how the learner trains, and how both get started.
02 Master Capture & Scoring
Not every clip a master films gets used. The AI learns well only from well-shot footage. So every take is scored for quality automatically, and the score sets how much it counts — this scoring system is already implemented and running inside the app.
From filming to training data
01 Filming The master films the standard process from a Galaxy XR first-person view, in 10–60 minute units.
02 Quality scoring The AI checks the footage on three criteria and scores whether it is fit to learn from.
03 Acceptance ratio The score decides how much of this take counts. A low score doesn't count.
04 Accumulation Once accepted footage reaches the per-dish goal, that dish's training data is complete.
What the scoring looks at — three criteria
Hands in frame Weight · important
Are the chef's hands in the shot? It matters for learning the motions.
Know-how narration Weight · most important
Does the chef explain why they do each step? This is 'transfer,' not just video — so it weighs the most.
A low score doesn't count
A high enough score makes the take count fully; a middling one counts partly; below the bar it doesn't count. That natural control — 'you filmed it but it didn't count' — nudges masters to film with care. Quality filters itself.
Score bands & acceptance — more detail
Each take is scored 0–100, and by band the acceptance ratio is 0% · partial · 100%. Accepted time = take length × acceptance ratio, reflected exactly in the per-dish goal. (The precise thresholds and weights are trade secrets and not disclosed.)
Today this runs as a first implementation keyed on take length; precise reading of the three criteria from the video itself will be advanced step by step as the vision AI is integrated.
03 The Learner's Journey
A learner may be cooking for the very first time. So learning splits in two — first grasp the dish's concept in your head (theory), and only once that's mastered do you move on to actually making it (practice). Both stages count by 'did you really learn it,' not 'how many times.'
0
Concept lectures
learned & scored by dialogue0
Accepted practice
to master the dish0 grades
Deviation from master
A · B · C · DMaster theory, then practice — the learning cycle
Theory
Theory — concept lectures Learn the dish's ingredients, tools and key points by voice — but not by just leaving it playing: you learn in conversation with the AI. The AI gauges your understanding from the dialogue, and if there's been little back-and-forth it asks questions at the end to check whether you truly grasped the theory, then scores it. Only past a certain level do you reach 'concept mastery' and move on to practice.
Practice
Practice — live dialogue learning The AI teaches by conversation, seeing the learner's live view. Each session is scored on engagement and comprehension, while also watching how close it is to the master (the deviation). Once enough sessions count, you reach 'practice mastery' of the dish.
How close to the master — deviation grade
A almost identical
B close to master
C noticeable gap
D large gap — keep at it
What the learner makes is compared with the master's standard and reported as one of four grades for how close it is.
Scoring & acceptance — more detail
Each session is scored 0–100, and the band decides how much it counts (none / partial / full). The deviation grade is also placed on a 0–100 scale. The precise thresholds and weights are not disclosed.
Today a dialogue-centered first-stage evaluation runs; evaluation that also reads precise alignment with the master's taste (across several flavor axes) will be added step by step as the camera and vision AI are integrated. As general AI models improve, our cooking-transfer AI is planned to be upgraded periodically too.
04 Getting Started (Master & Learner)
Masters and learners use the same app but take different paths — masters to 'leave' a dish, learners to 'learn' it. Here's how each track starts, with the real app screens. Use the selector just below to view the flow for whichever side you want.
- 1
Contract · invite → enter
Beyond an expression of interest, the target masters are brands that have signed a formal contract and are about to launch in the U.S. Contracted masters get an invite; after login they enter by choosing 'master chef' (language is chosen on the first screen).
- 2
Only your own brand
Select your brand from the list — you can enter only your own, the one authorized by invite. (Setup such as per-brand dish registration is done by the operator on the admin page, so the chef can focus solely on filming.)
- 3
Per-dish filming status
On the brand home, see each dish's filming progress (how much of the target is filled) and pick a dish to film.
- 4
Pre-filming briefing
Tapping 'Start filming' doesn't begin right away — first the basic concepts, evaluation criteria and cautions are shown as text. Then, after putting on Galaxy XR, you press start to begin filming in earnest.
- 5
Manage by sorting
When there are many dishes, the sort feature lets you decide as you go — continue your most recent take, wrap up a nearly-complete dish, or attend to the slowest one.
- 6
Review filming history
Look back over past takes by dish and time. See which dish on which date scored what, and review — by recency whether quality is improving, by score which dishes are weak — to apply to the next take.
Because learners are mostly based in the U.S., the app screen samples are shown in English; the language can be switched separately — to Korean and others — on the app's first screen.
- 1
Invite → enter & pick role
Log in with the invite from the master and a role-select screen appears. Choosing 'Learner' sends you down the learner path (no invite, no sign-up). On this first screen you can also pick the language — Korean, English, and so on.
- 2
Find your brand
Learners see a list of brands and go into the one they're opening. Only invite-authorized brands open; others can't be accessed.
- 3
Per-dish learning status
The brand home shows each dish's progress (not started · learning · done) at a glance. Only scores, grades and counts are kept — the original audio is never stored.
- 4
Review your history
Scroll the same screen down for your learning history. Sort by most recent or by score (high/low) to review whether your scores are improving and which dishes are weak.
- 5
Theory — in conversation with AI
The theory lesson starts with the dish's core concepts provided as text. The AI doesn't just read the text aloud — it maximizes the lesson's effect by exchanging questions and dialogue with the learner.
- 6
Finish theory to unlock practice
If you try to start hands-on practice before finishing theory, a notice appears and blocks you. Only after truly grasping the concepts can you move on to practice.
- 7
Practice — with Galaxy XR
For practice you wear Galaxy XR. The dark lens is only the outside — passthrough cameras show the real kitchen clearly, and its weight balance and field of view are stable enough to cook in. The AI sees that same view and teaches by live conversation, telling you in grades how much your cooking differs from the master chef's pattern.
