What's new

We spent the last week on speed. Images are now sized and cached once instead of being re-processed every place they appear, so the gallery and your creations come up quicker.

The app also warms up the studios and models you're most likely to open, and the studio preview stays ready after the first time you open it — so it pops up instantly instead of making you wait.

You'll feel it most when you're flicking through studios or browsing your gallery. Less waiting, more shooting.

If you use Colabz in light mode, the editing tools used to stay stuck in dark. They now follow your theme everywhere, so the whole app looks consistent.

We also fixed a small detour: when you erase, "choose from library" now opens a picker right inside the tool instead of sending you off to Assets and back.

Over the last two weeks we added 25 new studios to your gallery. Each one is a finished look you drop a product into and shoot: the lighting, surface, palette, and framing are already set, so there is no prompt to write and nothing to configure.

The range is wide on purpose. Warm leather ateliers like Maison Cuir and The Cognac Hour, bright poolside sets, a terracotta arch, athletic blues, icy auroras, and candy soft pinks. Pick the one that fits the product and generate.

They sit alongside the studios you already use. Open any of them to start a shoot.

May was one of our busiest months. The short version of what landed:

New tools

• AI Photoshoot: pick a category and get studio-grade product photos, no prompting needed.
• Studio Builder: build and save your own studios, with your brand baked in.
• Virtual try-on: a model plus your clothing item, composited into a clean studio shot.
• Reference to image: a model and a scene in, a six-shot editorial grid out.
• Model swap: now its own app, not a buried modal.

A smoother workflow

• Every generation and edit saves to Assets automatically, and you can star your favorites.
• A collapsible creations sidebar on the generate page, with faster switching between projects.
• A prompt enhance toggle, and a redesigned Products page with AI autofill.

Faster and more reliable

• Galleries load noticeably faster: the Assets and Faces grids are lighter and load images as you scroll instead of all at once.
• Snappier start: your credits and profile appear instantly, with no flash of empty or zero state.
• Quicker switching between projects and creations.
• Faster, more reliable downloads across the app.
• Plus a long list of smaller speed and stability fixes.

More coming in June.

The hardest part of building a feed used to be the location. You can shoot a model in a studio in an hour. Getting that same model into a Lisbon balcony at golden hour, then a tiled Marrakech courtyard, then a neon Tokyo alley, takes weeks and a flight budget. Reference-to-Image collapses that into one upload.

How it works

Pick one of our models or upload your own. Drop a reference image of any scene, a balcony, a kitchen, a beach, an alley, a hotel lobby. A vision model reads the location: light direction, palette, hero features, time of day. A director then writes six distinct story beats tailored to that specific place. Six parallel renders place your model inside the scene, matching the light, the depth, and the edges so the composite reads as one shoot.

What you get back

• Six editorial frames in a 3x2 grid, each a different beat of the same story.
• Aspect ratio auto-matches your reference, so vertical scenes stay vertical and wide scenes stay wide.
• Location-only references. Anyone in the reference image is excluded from the final frame, only your model appears.

What it is good for

• Campaign mood reels before you commit to a location.
• Lookbooks with a sense of place when you only have studio shots and a Pinterest board.
• Social grids that look like a real trip without booking one.

Cost

36 credits per generation, six shots, six credits each.

Sample shoots are expensive. Booking a model, renting a studio, steaming garments, paying a photographer, then waiting two weeks for edits, all to find out the top works but the pants do not. Virtual Try-On collapses that loop into a single shot.

How it works

Pick one of our models or upload your own. Drop in up to three garments: a top, pants, shoes, a full look, whatever you have flat-lay or on a hanger. We composite them onto the model on a clean white studio backdrop with consistent lighting, so the result reads as one shoot, not a collage.

What it is good for

• Product pages for lookbooks and PDPs when you only have flat-lays.
• Pre-sample reviews to see how a piece sits before you order the sample.
• Marketing assets when you need a clean editorial without a production day.

Cost

10 credits per try-on. Run it as many times as you want to swap models, swap pieces, or iterate on a look.

Most people building product pages don't want to write prompts. They want photos.

Drop one product image into Shots and you get four directed photos back: different angle, lighting setup, and composition for each. Behind the scenes it's one parent generation with four children, so the variants stay linked and you can re-run the set.

Best for products with a clear hero angle. If you want maximum control, the regular generator is still where you'll spend most of your time.

Library always felt like a separate place you had to remember to visit. Assets is the same surface, renamed, but it now materializes itself: every generation and edit shows up without you doing anything.

Favorites came along with the rename. Anyone in the workspace can star or unstar an asset, and the favorite is visible to everyone on the team. Useful for marking final picks before export.

Smoothing skin in AI tools usually flattens it into plastic. The Skin Enhancer takes a different angle: three opinionated modes that preserve identity and add back the textures most retouch tools strip away.

• Soft: editorial look, even tone, light smoothing.
• Realistic: balanced retouch, keeps pores and micro-detail.
• Imperfect: documentary feel, preserves natural texture and asymmetry.

Works on portraits and on-model product shots.

A text editor inside the modify flow with a small library of curated templates. Each template defines its own canvas size and layer arrangement, so the typography and spacing are already tuned. You just write the copy.

Useful for sale tags, product cards, drop announcements, and social banners.

A new documentation surface with category sidebar, markdown articles, and a media library bound to docs. Every article is chunked into a vector index, so the AI chat inside docs can answer specific questions instead of pointing you to a page.

Categories are editable without code changes, so the structure can evolve as we add tools.