Skip to content
Latentshoot
Menu

AI tools

AI tools for going through a gallery and editing a photograph

Analyse goes through a gallery for closed eyes and soft frames. Your client’s notes can be sorted into a list of what to change, and Draft a fix proposes an edit for each photograph. In Develop, a model can outline an area, redraw a patch you painted, extend a frame or match a reference. Nothing runs until you press it, the team owner has agreed and the gallery’s own switch is on.

Free includes 30 AI credits and 200 photographs to analyse, once, when you publish your first gallery.

The Photographs tab of a gallery after Analyse has run. A panel reads “All 26 photographs read” with three figures: 4 to look at again, 8 set-ups suggested and 9 tags suggested, each with a link to review them. Below, the Outerwear set of seven photographs of coats with filters for All, Picked, To judge and Rejected, and one frame of a man in sunglasses marked “Eyes ?”.
A gallery the day after Analyse ran. The man in the fourth frame is wearing sunglasses, so his eyes could not be read and the mark shows a question mark.

Analyse

Three questions about each frame

Press Analyse on a gallery or on one set and choose what to look for. A photograph that has been read is not sent again for the others unless you edit it. Set-ups on its own reads three frames from each burst, so choosing the others later sends the rest.

Tick Mark closed eyes and missed focus and each frame where somebody’s eyes are shut, or the subject is not sharp, is marked on its thumbnail. Where the model cannot tell, because of sunglasses, a face turned away or a portrait at f/1.4 that is meant to be soft, the mark has a question mark and is kept apart from the definite ones. It never rejects a photograph, and which of six nearly identical frames is the best stays your call.

With Suggest a set-up for each frame, Latentshoot puts each photograph beside the shot it most likely came from, using when it was taken and what is in it. It reads three frames from each burst and suggests one set-up for the whole burst. Nothing counts until you confirm it on the Set-ups tab.

Suggest tags offers only words already on your shot list, from its props, looks, groups and tags. You add each one to the photographs it fits, or skip it.

Your client never sees any of it. How a gallery works.

Client notes

A list of what the client asked for, and a draft for each photograph

Sort the notes, on a gallery’s Client tab, files each of your client’s notes under what it asks for, such as a crop, a colour change or something taken out, and opens Develop at the tool that answers it. It sends the words of each note and the photograph’s filename as you uploaded it, and no photographs and no names. Each photograph with notes counts once against your photographs to analyse.

Draft a fix reads every note on a photograph and proposes one edit: settings, a crop, or a change to take something out. It sends the picture your client saw, made smaller, with the notes and the areas they marked. A draft uses 2 AI credits a photograph, and nothing becomes a version until you keep it in Develop. If it proposes a change, you check the change before it is sent, and sending it uses a further 10 credits. How the notes list works.

The Client tab, headed Client notes: “14 notes sorted. 8 photographs have a request.”, marked AI, above 1 of 8 photographs done, 2 drafts to review and 2 changes to check, with buttons Review 2 drafts, Check 2 changes and Send the changes. Filters follow for All, Crop, Colour, Take out and Other frame, then one row per photograph with each note, who wrote it and what it asks for.
Fourteen notes from two people at the client, sorted by what each one asks for.

In Develop

Four tools in Develop

Type “the sky” into Select from a description and the mask is outlined for you. The sliders inside it are still yours, so nothing in the picture is generated.

With Change, in the Retouch tab, you paint over a wire or a sign and say what should change. Only the painted area is redrawn, and it sits over the photograph like a patch you can switch off or delete. Dust and most wires need no model: Heal and clone copies a clean part of the photograph over them, for no credits.

Extend, beside it, generates room round the photograph, in a shape you choose (Square, 4:5, 5:4, 2:3, 3:2, 9:16, 16:9) or edge by edge. The photograph itself is kept pixel for pixel.

Match a reference takes a photograph with the colour and tone you are after and suggests settings to match it. You choose which to keep, and can save them as a preset for as many photographs as you like.

More on Develop

Develop with the Retouch tab open, Change closed and Extend open. The photograph of a woman walking into the sea sits in the middle of a wider 16:9 frame, with the new area either side hatched and labelled New area, and the size reading 2,134 × 1,200 px. In the panel, a grid of shapes — Same shape, Square, 4:5, 5:4, 2:3, 3:2, 9:16 and 16:9, with 16:9 chosen — and Custom below.
Extending to 16:9, set up and not yet sent. The hatched margins are the only part a model will fill; the photograph in the middle is never redrawn.

Your photographs

Nothing is sent until you say so

Every control that sends a photograph has a small AI mark, so you can see which ones do before you press. The privacy policy names each service.

  • The team owner agrees once, from the first tool that asks or under Account, in AI permissions, and can withdraw it there.
  • Each gallery has its own switch under Settings, and it is off on every new gallery.
  • What goes is a smaller copy of the photograph, or of the part around your paint, with your words where a tool asks for them.
  • Draft a fix sends the picture your client saw, made smaller, with their notes and the areas they marked. Sorting their notes sends the words and the filename, and no picture.
  • Your shot list and your client’s details are never sent. Analyse counts the people in a frame and describes nobody.
  • It goes through Netlify’s AI Gateway to services that keep no copy of what they are sent.

Plans

What each plan includes

Extending a frame uses 25 AI credits, a change 10, a match 6, and a draft 2 for each photograph, plus a change’s 10 if you send a change it proposes. Analyse and sorting a client’s notes both count against your photographs to analyse, selecting from a description has a figure of its own, and none of the three uses credits.

Free

AI credits
30, once
Photographs to analyse
200, once
Selections from a description
Not included

Pro

AI credits
250 a month
Photographs to analyse
3,000 a month
Selections from a description
300 a month

Studio

AI credits
1,000 a month
Photographs to analyse
15,000 a month
Selections from a description
1,500 a month

Free’s figures are given once, when you publish your first gallery. On Pro and Studio you can buy 200 more credits for $10 or 1,000 more credits for $40, and they last 12 months. Monthly credits that go unused do not carry over. The full table is on the pricing page.

Where it stops

What the AI tools do not do

Nothing here writes over the file you uploaded. A change, a selection or an extended frame is kept beside the photograph, and deleting it leaves the frame as it was.

Analyse never picks or rejects a frame, and a suggested set-up or tag counts for nothing until you accept it. A sorted note is a suggestion read from the words alone, and a draft is a proposal you keep or skip.

A change is a model redrawing the area you painted, so look at it at full size before it reaches a client. Skin work, and anything a client will study closely, still belongs with whoever retouches for you now.

An extended frame is softer close up. The new area is generated at a lower resolution than your photograph and enlarged to fit, which suits a banner or a post better than a print.

An exported file with a change or an extended frame in it says, in its own details, that an image model made part of it. Instagram, Google and other sites can read that and label the picture as edited with AI.

Bursts, the camera clock correction on Coverage and the ways to catch up on set are not AI tools, and neither are Heal and clone or Your usual settings in Develop. They send nothing to a model and use no credits.

Questions

About the AI tools

Will Analyse reject photographs for me?
No. It marks the frames where it found closed eyes or a missed focus, and the frames where it could not tell. Picking and rejecting stay yours, with the same keys as Lightroom: P to pick, X to reject.
Does the model see my shot list or my client’s name?
No. Analyse sends the photograph and nothing else, and Latentshoot compares what comes back with your shot list itself. A renamed shot does not need the gallery read again.
Can I keep one client’s photographs away from it entirely?
Yes. Each gallery has its own switch under Settings, “Allow this gallery’s photographs and notes to be sent out”, and it is off on every new gallery. Leave it off and nothing in that gallery is ever sent, whatever the rest of the team does.
Does my client see a draft?
No. A draft stays on your screen until you keep it as a new version in Develop, and your client sees that version only once you show it in the gallery. They never see the sorted list of their notes either.
What happens to my credits if a job fails?
A job that no model answered gives its credits back in full, and a photograph Analyse could not read is not counted.
Is any of it on the free plan?
Yes, once. When you publish your first gallery, the free plan is given 30 AI credits and 200 photographs to analyse. Gradients and brush masks, which send nothing, are on every plan.

Try it on your first gallery.

Plan a shoot on the free plan, publish the photographs, and Analyse the first 200 of them.