Skip to main content
Next

Your taste

The nine rules are the same for everybody: a floor is a floor. Your taste is the other half — what you keep, learned from the decisions you have already made. Read this page when a mark says a lean or a bet and you want to know where it came from.

Bets and leans​

Two cells side by side: a bet on a solid disc and a lean as a hollow mark

On every undecided frame outside a set that no rule spoke about, Vixlo works out how likely you are to keep a frame like it, and marks the ends of the card:

  • A bet is 85% or better, or 15% and below. It is drawn as a full mark on a solid dark disc washed in its own ink.
  • A lean is drawn hollow, exactly as a floor's Reject or the burst's take is, so hover the mark to tell: its tooltip says a lean. There are two ways to earn one: 65% or better (or 35% and below), or a place in the top quarter of this card's own scores while above half — the bottom quarter while below it. The card's own quarters can only add a mark, never a bet.
  • Anything else is nothing at all, which is the app being able to say it does not know. When fewer than eight frames are left for your taste to judge — undecided, outside a set, and not already marked by a rule — there is no card to compare against, and only the bands speak.

Both keep the dash, because both are still only proposals. A lean says why on the reasons card — In your taste: you kept 4 of 5 like this — a subject this sharp — and waits for you; once accepted, one undo takes it back like any other decision. A group, session, stack or whole-card Accept leaves it alone; it is written by ↩ on the frame, by ⌘↩ in Single, or by Take all or Reject all beside an open proposal pass. See Accepting, or deciding for yourself. A lean towards Reject on a frame of a burst can also make a proposed pair in Compare with the burst's take, without the rest of the burst being proposed for the bin.

Your taste's row in the criteria pane: the Leans switch, the count line and what it has noticed

The Leans switch is on your taste's row in the Proposals and criteria pane, with a live count under it: Marks 41 frames of this card — 6 bets · 35 leans. Switched off, the line says what switching it on would mark, counted rather than estimated. Off leaves every floor, every burst take and the withholding below exactly as they were. The row also says Learning from your decisions has its own switch below. The two switches answer different questions — whether the app may learn from you, and whether what it knows may mark a frame — but Leans only matters while Learn from my decisions is on. Switch learning off and your taste leaves the proposals entirely, whatever Leans says: no bets, no leans, no choosing within a burst and no held-back Rejects, and its row reads switched off.

Before you have decided anything​

A fresh library has no evidence in it, so Vixlo starts from a small fixed table written down from what each measurement means: subject sharpness counts for more than frame sharpness; a well-captured face counts; a turned-away face counts against; a frame further down its burst counts against, and so does the second of a look-alike pair; a frame brighter than this card's typical exposure counts against, because nothing recovers a blown highlight; somebody you have named counts for a good deal. Two things count for nothing at all — how many faces, and the hour of day — because photographers want opposite things from both. That table was never tuned to anybody's decisions, and your own can still teach it either way.

Everything is judged against the library it is in: "sharp" means sharp for this card, so a shoot of soft available-light frames is not marked down for being one. A library whose frames all measure the same has nothing to compare against, and the app says nothing about any of them. As your decisions pile up, "sharp" comes to mean sharp across everything you have culled, and it moves over gradually, so a handful of decisions never re-marks a whole card.

The three things it may do to a proposal​

Your taste is allowed to touch a proposal in exactly three ways.

  1. Lean on a frame no rule spoke about — the bets and leans above. A frame you have decided, a frame in a set, a frame a rule has already spoken for, and a frame whose Reject your taste held back never get one.
  2. Choose which frame of a burst is proposed as the take, but only among frames the sharpness cannot really separate — within 6 points on the 0–100 scale — and only when your taste puts one of them clearly ahead. A visibly softer frame never displaces a sharp one, and a close call leaves the measurement's answer alone.
  3. Hold back a Reject. When your taste scores a frame at about seven in ten kept or better, and the rule proposing the bin is not a floor, the app says nothing about that frame at all. It does not become a Take; it becomes silence, and the reasons card says so: The look-alike rule proposed the bin; your taste held it back — you keep 7 in 10 like this. A floor always stands: the sharpness floor, the exposure floor and accidents are never withheld, because whether there is an image in the file is not a matter of taste.

Where a floor has proposed the bin, that is what is proposed, and the taste does not get a word in beside it.