Changelog

2026-09-18

New pricing: plans by storage, free seats, credits for generation

Three plans replace per-seat pricing: Solo at $39 a month with 1 TB, Pro at $99 with 5 TB, and Agency at $199 with 15 TB. They differ by storage only, and they’re billed monthly.
Seats are free and unlimited on every plan. Invite your whole team and the bill doesn’t move. Client reviewers stay free and still don’t need an account.
Generation runs on prepaid credits, for images as well as video. One credit is one cent, every model shows its cost in credits before you run it, and a generation draws down what it costs to make, doubled.
Each plan comes with credits to start — 500, 1,000, or 2,000 — added once when the plan first activates.
Credit packs now start at $10, alongside $25, $50, and $100. Credits never expire, and a failed generation isn’t charged.
Past your plan, storage is $20 per TB per month. Going over never blocks uploads or deletes anything, and nothing is added to your bill without the team getting in touch first. Owners and admins get an email at 80%.
If credits reach zero, generation pauses and everything else keeps working. If a plan ends, generation and uploads pause, nothing is deleted, and unused credits stay put.
The full picture is on the Billing page.

2026-08-12

Upscale and Remove background

Upscale adds resolution and detail to any image; Remove background cuts the subject out onto a transparent background. One image, one button, no prompt and no settings.
Both live under the new Tools section in the sidebar alongside Reframe, and you can start either from an image you’re already looking at in your Library.
Results save to your Library next to the original and land in the same collections. The original is never touched.
A before/after view shows what changed, at 100% zoom — an upscale is invisible at screen size.
Drag a file in from your desktop and it’s added to your Library first, so nothing ends up floating outside it.
Running the same tool on the same image twice shows you the result you already have instead of charging again.
Free to use, like generating images.

Upscale and Remove background from connected AI tools

Connected AI tools can now upscale an image or remove its background — point one at an image in your Library and the result saves back beside it, same as if you’d done it in the studio.
They can find the image first with the existing search, so you can just say “remove the background from that product shot”.
Everything the studio refuses, a connected tool refuses too — including upscaling a cutout, which would fill the transparency back in with white.

Remove members, rename brands

Studio owners and admins can remove someone from the studio via the ⋯ menu next to a member on Settings → Members. They lose access to the studio and every brand in it; their generations and uploads stay where they are.
You can’t remove yourself, and only an owner can remove another owner — so a studio can never end up with nobody in charge.
Brands can be renamed: the Kit tab has a Name field. The new name shows everywhere, and the brand’s link stays the same so existing bookmarks keep working.

Google’s image models, brought up to date

The Google card is now Gemini, with three tiers: Flash for everyday work, Pro for studio-grade 4K, and Lite for cheap, fast iteration.
Google switched Imagen 4 off on August 17. Anything you made with it still lives in your Library, and re-running an old Imagen recipe quietly uses the closest Gemini model instead of failing.
Wide and tall crops work properly now: 21:9, 4:5, 2:3 and the rest used to be offered and then rejected. Flash and Lite also take banner shapes down to 8:1.
Pro renders true 4K; Flash goes up to 4K or down to a fast 512px draft.
Seed, negative prompt, and watermark controls no longer appear on Gemini models — they were doing nothing.
Downloads now carry the right file type. Every generated image used to be saved as .png regardless of what it actually was.

2026-08-09

Connect your AI tools to the studio

MCP-capable AI tools can now connect directly to your studio — they can search your Library, read your brand kits, browse models, and generate images that land in your Library like any other generation.
Create a key under Connected AI Tools in your account settings. You’re shown it once, it works only for the studio you’re in when you create it, and you can revoke it at any time.
Everything a connected tool does respects your existing brand permissions, billing status, and daily limits.

New Billing tab in Settings: subscribe your studio per creator seat and you’re unlocked the moment payment clears.
Only people who create count as seats — client reviewers and shared campaign pages stay free and unlimited.
Card changes, invoices, plan switches, and cancellation are all self-serve, and adding or removing teammates adjusts your bill automatically.
Video generation runs on prepaid credits with a published rate card. Credits never expire.
Your storage use now shows on the billing page, and going over never blocks or deletes anything.

Workflow pages: shape the LoRA stack and tune every setting

Workflow pages show the LoRA stack the workflow ships with — reorder, adjust strength, remove, or add compatible LoRAs from the catalog before you generate, and reset to the defaults any time.
Sampler, steps, guidance and the rest moved into a collapsed Advanced settings section that starts from the workflow’s tuned defaults and shows how many settings you’ve changed.
Studio owners and admins can add or remove workflows for their studio right on the Workflows page.
Studio operators can tune each workflow’s default LoRA stack and settings centrally; new sessions pick the changes up immediately.
Adult LoRAs stay invisible in studios that haven’t enabled mature content — in the catalog, on workflow pages, and at generation time.

Pick your studio’s look — color schemes and light mode

Four new color schemes join the default look: Starry Night, Kelvin, Blueprint, and Arc — pick one under Preferences on your profile.
Alchemy, the bright magical-indigo the studio launched with, is back as a scheme.
Dracula joins too, built from the official palette, with the full chart cast.
Every scheme comes in both light and dark, with a new Light / Dark / System toggle next to the picker.
Your choice applies instantly, sticks on this device, and doesn’t touch anyone else’s view.

2026-08-08

Slack review threads now sync both ways

Comments on an asset mirror into the Slack thread under its review message — with your name and avatar, so clients read a conversation, not a bot feed.
Replies in the Slack thread land back on the asset in the studio, attributed to whoever wrote them — including clients on Slack Connect who don’t have a studio account.
Reactions, edits, and deletes follow along in both directions, and Slack replies ring the submitter’s notification bell.

Folder import got a whole new home

Importing a folder now takes over the screen instead of squeezing into a small dialog — a rail on the left keeps the destination, progress, and what draft vs approved means in view the whole way through.
The first step shows the actual files you dropped and a collapsible walkthrough of what happens next.
Progress feels alive: a big live count, a stage-by-stage timeline, and your own files lighting up as each one is processed — plus honest screens when something waits, disconnects, or fails.
Review is one collection at a time: pick a collection on the left, see its photos and tags on the right, and choose Approve now or Leave as draft.
Run several folder imports at once — minimized imports stack as cards in the corner, each with live progress and a Review button when ready.
Closed the window mid-review? A banner in the Library (and the bell notification) brings the exact review screen back.
Imports stopped writing junk: the dropped folder’s machine name is no longer added as a tag on every file, and collections get a searchable title from what’s actually in the photos.
There’s now a visible Import a folder button in the Library toolbar — no more knowing to drag a folder.

Workflows are back — and now curated per studio

The Workflows section is live again: tuned generation pipelines, each with its own page and settings.
Workflows are enabled per studio from a central registry — your studio only sees the workflows added for it.
New studios start with a clean slate; ask your administrator to add workflows from the registry.

”Reject” is now “Request changes”

Review actions across the studio say Request changes instead of Reject — clearer that the work comes back for revision, not a dead end. Its status reads Changes requested.
Same everywhere: the review queue, bulk actions, filters, status badges, and the Slack buttons.

Studios are now invite-only

New accounts no longer get an automatic starter studio — you land directly in the studio you were invited to.
Accept an invite to a team studio and the untouched starter studio from your signup is cleaned up automatically. Anything you actually used is never touched.

2026-08-07

Your studio now lives in a chip at the top of the sidebar — click it to switch studios or jump to Studio settings, Manage brands, and Usage.
The account menu at the bottom is down to the essentials: What’s New, Account settings, help, your XP, and log out.
Interface scale moved to Account settings under Preferences — same control, same per-device behavior.

Approve reviews right from Slack

Submitted-for-review messages in Slack carry Approve and Request changes buttons, with the asset image shown inline.
Request changes opens a dialog for your feedback note (required), and the submitter gets notified exactly as if you’d reviewed in-app.
After a decision the Slack message updates in place — who decided, and when, in your own timezone.

Connect your Slack account

Link Slack from your profile settings — studio Slack updates then @-mention you instead of just showing your name.
Your linked identity is also what lets reviews be handled straight from Slack.
Disconnect any time from the same place.

Studio wins in Slack

Pick a Slack channel in Studio settings and your studio’s fun moments post there automatically — level-ups, feats (including 7- and 30-day streaks), and new members joining.
On the 1st of each month, the studio’s top creator from last month gets a leaderboard shout-out in the same channel.
Off by default — nothing posts until an admin sets the channel.

2026-08-06

Slack notifications for asset reviews

Connect a Slack channel and your studio posts a message whenever an asset is submitted for review, approved, or sent back with change requests — with the asset image attached and a link that opens it in the studio.
Each brand can post to its own Slack channel — set it in the brand kit page, ideal for per-client channels. Brands without one use the studio default.

Folder imports show live progress again

Importing a folder shows real progress the whole way through, instead of sitting at 0 while the import quietly finished in the background.
Type your own tags when importing — the tags field takes free text, alongside the folder suggestion and common-tag chips.

2026-07-03

Fixed image edits failing with a false “flagged as sensitive” error

Image generations — especially edits that use a reference image, like swapping a background — no longer fail intermittently with a “flagged as sensitive” error on ordinary prompts.
The overly-strict filter behind those false alarms is gone.

2026-06-27

Browse your Library by date

Your Library groups by date automatically — newest first under Today, Yesterday, Earlier this week, and month headers, plus a Jump to rail to skip to any period.
Switch between Uploaded and Captured to organize by when files were added or when photos were actually taken.
Click Timeline to reveal an upload-volume chart — drag across it, or pick Last 7 days / Last 30 days / This month, to narrow the grid to a date range alongside your other filters.
Time-filtered views are shareable — the date range lives in the URL.

Upload assets from anywhere with ⌘K

Open the command palette (⌘K) and pick Upload assets to start an upload from any page.
Choose files for a plain upload into your active brand, or choose a folder to run the organize-it-for-me import.

2026-06-26

We’re now called Cauldron

The studio is now called Cauldron — the new name shows on the sign-in screen, the browser tab, the About dialog, and the emails we send you.
Same studio, same sign-in, same library, brands, and history — only the name changed.

2026-06-24

Drop a whole folder into your Library and it organizes itself

Drag a folder onto your Library and import every file into one brand in a single pass, instead of uploading and tagging batch by batch.
Each image and video is auto-tagged and grouped into a handful of collections, with anything uncertain set aside in a Needs-review group — and no folders are ever created.
Pick where it lands by name: the destination step shows your brand (and optional campaign) clearly, so you always know the access boundary before importing.
Review the result a collection at a time — rename, fix tags, include or skip a group.
Long imports keep running in the background: minimize to a card, keep working, and get notified when it’s ready. Re-dropping the same folder fills only the gaps.

2026-06-21

The Assistant can search your library and look things up

Ask the Assistant to “find my…” or “search my library” and it pulls from your assets, uploaded references, collections, brews, and LoRAs.
It can look up your brands and their kits, saved brews, and LoRAs on request.
Ask what a model costs or what it’s good for and the Assistant can explain and compare models before you generate.
Ask it to “polish” or “tighten” a prompt and it rewrites it for you.

The Assistant can reframe your work and show it to you

The Assistant can reframe your assets into every social format and notifies you when the crops are ready.
Ask to “show me my last creation” and it displays the actual image inline, with the usual open-full-size and download controls.
Ask for “my recent work” and it reaches back past today, so “show me what I made” finds it even if you haven’t generated since yesterday.

Dark mode: muted text is easier to read

Secondary and helper text in dark mode is brighter, clearing WCAG AA contrast with room to spare.

2026-06-20

Reframe, redesigned — a live preview and aspect ratios for any size

Reframe opens as a full composer: controls on the left, a live preview on the right that updates as you toggle formats and fit.
Pick by social format and by raw aspect ratio (1:1, 4:5, 9:16, 16:9, 2:3, 3:2, 1.91:1, 21:9), so web banners and print sizes aren’t stuck behind platform names.
Formats that share a ratio make a single crop tagged for all of them — Story, Reel, and TikTok become one 9:16 file, not three duplicates.
Reframing a collection? The preview steps through your images one by one so you can spot-check the batch.
Download the new crops as a zip from the result, or jump to them in your Library, already filtered and tagged by format.

Choose the Assistant’s chat model — and the model that makes your images

Two clearly-labeled model menus: a Chat model in the message box that powers the conversation, and an Image model on the Generate card that makes the picture.
Pick a specific image model before you generate, or leave it on Auto. Your pick sticks for the rest of the conversation.

A calmer, clearer Assistant

While the Assistant is thinking or making an image, you see a live status with an elapsed timer instead of a bare spinner — and if it’s taking unusually long, it says so and offers to retry.
If a generation hits a snag, the chat shows a friendly explanation with a Try again button instead of a raw error code.
Replies break into proper paragraphs and wrap at a comfortable reading width. Headings, lists, quotes, and code blocks have clearer rhythm.
Sending a prompt jumps the chat to the newest message, and the view auto-scrolls as a reply streams — unless you’ve scrolled up to read.
Asking for several images in one message now shows all of them.

Rate the Assistant’s answers

Every Assistant reply and generated image has a quiet thumbs up / thumbs down.
Thumbs-down opens an optional note (Off-prompt, Wrong style, Bad quality, Not what I meant), or you can skip it.
Your ratings stick: reopen a conversation and the thumbs you gave are still there.

2026-06-19

New Reframe tool: crop and resize images into Instagram, Story · Reel · TikTok, and YouTube formats in one go — pick a collection, several images, or a single image.
Choose Smart crop (fills the frame, tracks the subject), Fit whole image (pads so nothing is cut off), or Auto (both).
Start it from the sidebar, a collection’s ⋯ menu, the library multi-select bar, an image’s Tools panel, or by asking the Assistant.
Watch live progress while it runs, and get a notification when it finishes.
Outputs land back in the same collection, tagged by format — no new folders.

2026-06-12

Cleaner thumbnails

Thumbnails in the Library and Gallery are uncluttered: the type, status, brand, and creator badges fade in when you hover a tile instead of always sitting on top of the artwork.

2026-06-06

Notifications you can actually scan — and act on

Every notification leads with a thumbnail of the asset it’s about.
Click a review notification and the exact asset opens instantly, right where you are. Approve or request changes without leaving the spot.
Mentions and replies open straight to the asset’s comment thread, focused on the message you were tagged in.
New Needs you tab gathers just the items waiting on you — reviews to do, invites to answer, changes to make — with a count.
Notifications group by time (Today / This week / Earlier), and approvals read green while change requests read red.

A faster ⌘K

⌘K includes the Assistant, so you can jump straight to chat from anywhere.
Image and video results show a real thumbnail with the file name and its brand chip.
Commands split into Quick actions and Go to navigation, so the two kinds of results read as distinct.

2026-06-05

One consistent preview for everything you generate

Open a result in the Image, Video, or Workflow studio and it appears in the same rich preview you get everywhere else — a big view with zoom, prompt details, ratings, tags, and collections.
Flip through everything from your session with ← / → right inside the preview.
Regenerate is one click from the preview, and on images Use as input drops the result straight back into your composer.
Comment threads, downloads, and Find similar are available on freshly generated results too.

Open any asset in place

Pick an image from ⌘K search and it opens right where you are, in a full preview with all its details.
The same in-place preview opens from chat results, campaign pages, and profiles.

Bring your own image — and @-mention — into the Assistant

Drop a picture into the chat composer — upload, paste, or pick from your library — and the Assistant edits that image.
New Use from your library button opens your workspace images so you can iterate without re-uploading.
Type @ to reference an asset or a brand: @ an image to edit or build from it, @ a brand to keep the result on-brand.
Attach up to four images per message.

Image batches that all show up

Ask for more than one image and you get all of them — a batch of 2, 3, or 4 no longer quietly comes back with just one when the image service is busy.

2026-06-04

Redesigned Home — search, or describe what to create

Home opens with a personal greeting and a single bar — type to search everything in your studio, or describe something to make.
The bar opens the ⌘K palette: your matches appear as you type, with “Create an image of …” and “Create a video of …” right below.
Below the search bar, a Fresh from the cauldron feed shows your studio’s latest activity.
A Your momentum panel shows your pieces this week, day streak, level, monthly rank, and badges.
Tap the mic to dictate what you’re looking for, or the paperclip to attach an image and go straight to editing, animating, or generating from it.

Voice typing

Dictate instead of type — tap the mic in the prompt box and speak your idea; the words drop in at your cursor for you to review and edit before you generate. Nothing sends on its own.
Dictation knows your setup: the brand, model, and LoRAs you’ve picked are used to spell tricky names and trigger words correctly.
The same mic shows up in asset thread replies and in ⌘K search.
If a recording doesn’t go through, your audio is kept so you can retry with one tap.

The studio fits your screen

On large and ultrawide monitors, pages use the extra width instead of stranding everything in a narrow center column.
On big displays the whole interface scales up a notch automatically.
Choose your own interface scale — Auto, or 90% up to 125%. It’s remembered per device.
Reading and settings pages stay comfortably narrow on purpose.

A smarter Assistant

The Assistant writes its own rich, detailed prompts from your idea instead of echoing your words.
It can see the image it just made, so “make it warmer” or “now at night” iterates on that picture instead of starting over.
Voice dictation works in the chat composer.
Rename, archive, or delete any chat from the thread header, and your full history is a click away.

2026-06-03

Getting Started quest board

The first-run setup screen is now Getting Started at /getting-started. Old /welcome links redirect here.
Your four essentials — complete your profile, create a brand, upload an asset, and make your first piece — are four prominent cards with a progress tally.
Everything else moved to a separate, optional Bonus list that never counts against your setup.

Redesigned Brands page with Gallery and List views

The Brands page has two views you can switch between — a Gallery of brand cards and a compact List — and it remembers which one you prefer.
Each brand shows more at a glance: its color, your team’s avatars, your role, asset and campaign counts, and when it was last active.
A search box filters brands by name in either view, and List lets you sort by name or asset count.

2026-05-31

Flux 2 Pro is the new default image model

With the image model set to Auto, the studio uses Flux 2 Pro for everything — plain text-to-image, generating from reference images, and editing.
Flux 2 Pro can work from up to 8 reference images in a single generation.
The Flux card leads with Flux 2 Pro, with Klein (fast, low-cost, LoRA-friendly) and Kontext (editing) still selectable. Flux Pro 1.1 and Flux Dev are no longer offered for new generations.

The top of the left sidebar shows your studio’s name and logo — set them in Studio settings.
The account card at the bottom shows your display name instead of your email.

2026-05-30

Assistant moved into the main sidebar

Chat is now Assistant, a regular item in the left sidebar right below Home — your studio tools stay visible the whole time you chat.
Your chat history opens from a History button, with a search box, chats grouped by Today / Yesterday / Last week / Older, and a collapsible Archived section.

Pick your brand once, everywhere

Brand is now a global control in the top bar, next to the notifications bell — choose it once and it sticks as you move between the Library, the composer, and the gallery.
Your choice is remembered across page reloads.
Pick All brands to see everything at once, or a specific brand to focus.
The separate brand pickers inside the Library are gone — the top-bar control is the single place to switch brand.

Quick actions in ⌘K search

The ⌘K palette opens with a Quick actions section that leads with Create image and Edit image.
Navigation rows read as verbs — “Go to Library”, “Go to Gallery”.

2026-05-29

Opening a Library asset fills the screen with a large, centered viewer instead of a narrow side panel — the image shows full size on a dark stage, at its true shape.
Click the image to zoom to actual size, scroll to zoom, and drag to pan.
Step between assets with the arrow keys or on-screen arrows, and press Esc to close.
Everything from the old panel is still here — rename, rate, tag, collections, campaigns, pin to brand, Find similar, Use as input, Download, Delete, and the Thread tab.

Workflows now have dedicated pages

Each workflow opens on its own page with a description of what it does, how long it typically takes, the settings it exposes, and the reference images it needs.
Settings are tuned per workflow — you only see the controls that matter.
Workflows that need reference images won’t let you generate until you’ve added them.

Text, Hybrid, and Similar search ranking now live behind a single search-settings button that shows your current choice.
Pressing / still jumps your cursor into the Library search box.
The Add reference picker has a search box — search follows the tab you’re on, and History shows every match rather than just your most recent.

2026-05-28

Composer redesign — model first, simpler settings

The model picker sits at the top of the composer, above the prompt — pick what you’re generating with first, then describe it.
References are one tap away: a single Add reference button opens a unified picker where you can drag and drop, upload, or browse past uploads and your gallery.
Aspect ratio and image count stay in view; the rest tuck into an Advanced settings panel.
The same refresh landed in the video studio, with first-frame references.

Generate up to 4 at once — results on the same page

Results stream in right next to the composer instead of replacing it — your settings stay put and you can keep generating.
New image-count stepper (1–4) makes several variations in a single click.
Hover any image in the results feed to download it or use it as a reference, or click to open it full size.
Your recent generations are restored in the feed when you reload the page.

Find similar, and search by meaning

New Find similar action on any asset surfaces the closest visual matches in your library — good for spotting variants and near-duplicates across a shoot.
Works for both uploaded photos and AI generations.
The search bar has a Text / Hybrid / Similar toggle: Text matches exact words, Similar ranks by visual meaning, and Hybrid (the default) blends both.
If an asset is brand-new and still being indexed, Find similar tells you to check back shortly.

2026-05-27

A new way to create — dedicated Image, Edit, and Video studios

The Create button in the sidebar opens a menu: jump straight to a new image, an image edit, a new video, or the Assistant.
Each is its own focused workspace instead of one crowded page.
New Auto model setting picks the best model for your prompt automatically — or open Browse models to filter by provider, capability, and what each model is best for, with speed, max resolution, and reference support shown at a glance.
Brand kits, references, LoRAs, and every generation setting work exactly as before.

Library, now with nested collections

The Library has a left rail: browse your collections as a folder tree and click any collection to filter the grid.
Create, rename, move, and delete collections right from the rail, and nest them inside one another.
Deleting a collection removes it and everything nested inside it, but never touches the underlying assets.
The separate Assets and Collections sidebar items are now a single Library entry.
Gallery now shows only generated media — uploads and imports live in the Library.
Rate assets 1–5 stars, and filter by star rating or by source (generated / uploaded / imported).
Select multiple assets to tag, set status, move between brands, add to a collection, download as a zip, or delete.
Drag a selection of assets straight onto a collection in the rail to file them there.

A cleaner Home — now a launchpad

Home is built around what you want to make: a big search/command bar (⌘K), tool tiles, and your most-used Brews as one-click starting points.
The recent-activity feed, drafts, the review queue, and the weekly throughput card moved to Activity, Gallery, and Review.

Creators can start campaigns

Anyone with the Creator role on a brand can now create campaigns — previously only Brand Managers could.
Editing and deleting campaigns is still limited to Brand Managers and studio admins.

2026-05-26

No more blacked-out generations

The Flux, Kling, and Wan models moved to a host that doesn’t blank results when its content filter trips.
Same models, same prompts and reference images — the old provider would silently return a solid-black image.

2026-05-22

Brands page opens the real configurator

Clicking a brand on the Brands page opens its full configurator — color, brand kit, palette, prompt rules, members, and campaigns — instead of a stripped-down name-and-color popup.
Creating a brand drops you straight into its setup.

2026-05-21

Stuck brews clear themselves

Brews stranded “brewing” — usually when a generation was interrupted before it could start — now time out and clear instead of piling up in your queue.

2026-05-16

Search anything from anywhere with ⌘K

Press ⌘K (or Ctrl+K) from any page to open the search palette — find brands, brews, and assets by name or by meaning.
Type a word and get instant matches; assets and brews stream in as the search runs.
An empty palette shows your most recent pages, so jumping back is one keystroke.

Library search now finds assets by meaning

Search the library for what an asset looks like, not just what’s in its filename or tags — “sunset on a beach” surfaces the right images even if nothing says “sunset”.
Every asset gets a Find similar affordance.
Works on images and short videos; all of your existing assets get processed automatically.

2026-05-15

Invite teammates and switch between studios

New Settings → Members tab: invite people by email, change roles, toggle who can generate video, and resend or revoke pending invitations.
Invitation emails carry a secure 14-day link.
Strict email match on accept — if the invite went to one address and you’re signed in as another, the page tells you to switch accounts first.
A studio switcher lets you move between studios you belong to.

Assign brands when you invite someone

The invite dialog has an Add to brands picker — toggle any team brand and the invitee lands as a creator on each one the moment they accept.
Brand assignments show on each pending invitation row.

Edit your display name and avatar

Open your profile and hit Edit profile — the same dialog updates your display name and uploads a new avatar in one go.
Avatars accept PNG, JPEG, or WebP up to 2 MB, and the change shows up everywhere as soon as you save.

Notifications live in a bell in the top-right of every page header.
Queue collapses to a small “N brewing” pill next to + Create, and disappears when nothing is in flight. Press Q to toggle it from anywhere.
What’s New, Help & docs, and a new Keyboard shortcuts dialog moved into the user menu.
New Discover section at the bottom of the sidebar, with Leaderboard.

2026-05-13

Tags and collections in the library

Tags are first-class — each has a name, a color, and an optional category, and renaming a tag updates it everywhere it’s used.
Open any asset and tag it from the detail panel: type to find an existing tag, or press Enter on a new name to create one.
New Tag and Collection buttons in the bulk action bar — select multiple assets and add or remove from up to 200 at once.
Collections give you reusable, manually curated groups.
Two new admin pages: Settings → Tags (rename, recolor, recategorize, merge, delete) and Settings → Collections.

Consistent asset tiles, and a bulk bar that fits

Gallery, Library, campaign pages, your profile, and the studio overview share one tile — same layout, hover behavior, and metadata corners.
The bulk action bar no longer runs off the right edge: the four organize actions collapse into a single ⋯ Organize menu.
Disabled actions are dimmed but readable, and the tooltip tells you exactly why each one is unavailable.

2026-05-12

Sign-in emails match the studio, and tell you which host you’re signing in to, that the link expires, and that it’s one-time use.
After submitting your email you land on a branded page that echoes back the address, so you can spot a typo before hunting in your spam folder.
Gmail, Outlook, iCloud, Yahoo, and Proton users get a one-click link straight to their webmail.

Asset details and tiles work properly for uploads

Opening an uploaded image’s details no longer says “Generated with upload” — it reads Uploaded · date (or Imported · date).
Uploads show a File name block instead of a Prompt block.
Per-file upload rows have meaningful states: queued, uploading with real progress, done, or failed with a one-click retry.
Errors that aren’t worth retrying (file too big, wrong type) skip the retry button.

2026-05-09

Workflow picker and a cleaner Workflows page

New workflow picker in the composer: pick the model first, then the workflow.
Multi-Reference Edit landed as a new workflow — upload a scene image plus one or more reference products and it swaps in the references while preserving pose, lighting, and framing.
The reference-image affordance reads off the active workflow, so switching workflows updates the upload tile in place.
Every generated asset records which workflow built it.
Workflow cards are click-anywhere, with a filter bar (All / Text to Image / Image to Image) and counts.

2026-05-08

Generation queue — submit and move on

Submit a brew and the form clears immediately — no more waiting on a spinner before the next prompt.
New Queue page shows what’s brewing now and what finished in the last 24 hours, with thumbnails and live elapsed time.
The sidebar Queue link shows a live count of brews in flight.
Stuck brews recover automatically, and stale-looking ones dim and flag their elapsed time.

Workflows directory

New Workflows page in the sidebar. Browse every workflow the studio knows about, see its settings, and drop into the composer with the right configuration applied.
New Flux 2 Klein Edit workflow — image-to-image tuned for stronger reference adherence.

Asset details fixes

Klein Edit generations show the reference image you used in the asset details panel.
The Enhanced Prompt section is hidden when it matches your original prompt — no more confusing duplicate text.

2026-05-07

Bulk actions in Library and Gallery

Hover any card to reveal a checkbox — click to select, shift-click to extend a range, ⌘/Ctrl+A to select everything loaded, Esc to clear.
A floating action bar appears the moment you select something — submit, approve, request changes, archive, delete, move to brand, or assign campaigns to the whole batch.
Buttons that don’t apply to your selection dim with a count tooltip, so you know exactly what an action will affect.
Approved assets stay protected — bulk delete and bulk move skip them and tell you which ones need a fork.

Tag assets to campaigns from anywhere

The campaign picker appears in the review modal, so managers and creators can group assets while triaging.
After a generation finishes, a campaign picker shows up next to the brand and brew controls.
Only creators, brand managers, and studio admins see the picker.

2026-05-02

Activity feed

New Activity page in the sidebar — a real-time log of everything happening across your workspace.
Three tabs: For You (events relevant to you — your submissions, approvals, rejections, XP, badges), My Brands (all events on brands you belong to), and Workspace (every event across the studio, admins only).
Filter chips narrow the feed to specific event types: approvals, drafts, feats, or level-ups.
Time filter lets you scope to today, the past 7 days, 30 days, or all time.
Infinite scroll with cursor-based pagination — the feed stays fast as history grows.

Campaign tagging across the full workflow

Tag assets with brand campaigns at two entry points: the Generate page (at creation time) and the Library detail panel (after the fact). The Review modal shows campaign assignments read-only.
Campaign chips appear on asset cards in the Library so you can scan which initiative an asset belongs to without opening it.
Workspace teammates can now edit tags and campaign assignments on any shared asset — previously limited to the asset’s creator.

Public campaign galleries

Share a campaign’s approved assets with anyone — no login required.
Each campaign gets a shareable public URL at /brands/[slug]/campaigns/[campaign-slug]/public.
Only approved assets are visible. Draft, in-review, rejected, and archived assets stay private.
Brand managers control which campaigns have a public gallery.

Threads everywhere

Threads (introduced in Library on 2026-05-01) are now available in the Gallery and Review modal as well — the Thread tab appears on any asset wherever you open it.

2026-05-01

Now called OpenCauldron

The studio is officially OpenCauldron — the styleguide, sidebar branding, and downloaded filenames all carry the new name.
Saved files now download as opencauldron-{id}.webp / opencauldron-{id}-original.{ext} (was cauldron-…).
The seed brand previously named “Cauldron” is renamed to “OpenCauldron” in place — same brand, new label, no asset moves.

Inside the review modal, a horizontal filmstrip rail at the bottom now shows what’s coming next and what just passed — no more guessing where you are in the queue.
Click any thumbnail to jump to that asset; the active tile auto-scrolls into the center as you walk the queue with j/k.
Approved/rejected items stay in the strip, dimmed with a check or X marker, and j/k automatically skip over them so you only land on what still needs a decision.
Hovering a thumbnail shows the asset’s prompt as a tooltip — useful when 52–64px thumbs aren’t enough on their own.
Honors prefers-reduced-motion: the strip jumps instantly instead of smooth-scrolling when the OS preference is set.

Threads on every asset

Every Library asset now has a Thread tab — pin a conversation right next to the work itself, not in a separate Slack channel.
Messages stream in real time, so a teammate’s note shows up the moment they hit send.
React with any emoji, @-mention a workspace member, or reply to a specific message to keep tangents readable.
Paste or drag images, GIFs, and video clips straight into a message — they upload inline and render in the thread.
Attach an existing Library asset to a message with one click — the recipient sees a tactile asset card that opens the original.
Mention notifications drop into your bell with a deep-link that opens the asset and jumps to the message.

Downloads work again

Fixed Download in the Gallery and Library failing with a silent error on hosts where the storage bucket isn’t CORS-configured.
Files now stream through the app, so the browser always treats them as a download (with the right filename) instead of opening them in a new tab.

2026-04-30

Review queue opens into a gallery

Picking a brand on /review now shows a scannable gallery of pending assets — no more forced single-item modal flow.
Click any tile (or press Enter on it) to open the existing review modal at that asset.
Closing the modal returns you to the gallery; only the last asset reviewed pops back to the brand list.
Loading, empty, and error states get tasteful Card treatments — including a Try again button if the queue fetch fails.

Faster image previews + pick your download size

Library and Gallery full-size previews now load a compressed WebP version — typically 90%+ smaller than the original, way faster on slow connections.
Download button is now a split button: one tap grabs the smaller WebP for sharing, or open the menu to pick the original at full quality.
File sizes are shown right in the menu so you know what you’re getting.
Mobile gets a single full-width download button that always opens the menu — easier to tap.
Originals are never modified — the WebP is purely additive and silently falls back to the original if anything goes wrong.

Library, upgraded for teams

Workspace admins now see every brand’s assets in the Library — not just their own.
Brand filter shows all the brands you have access to, with personal libraries labelled by member name and avatar so multiple Personals are easy to tell apart.
Each thumbnail now shows a status chip (Draft, In review, Approved, Rejected, Archived) so you can scan workflow state without opening the asset.
A creator avatar in the bottom-right of every card shows who uploaded or generated it — hover for their name and the action.

Send feedback from Slack

Type /feedback in Slack to fire off a quick note, bug report, or Loom link straight to the team.
A short form keeps your feedback in context — no copy-paste, no lost threads.

2026-04-29

Prompt enhancer, redesigned

Enhance now lives inside the prompt box — one tap rewrites your prompt in place, no extra panel to expand.
Hit Cmd/Ctrl+E to rewrite without leaving the keyboard.
Don’t love the rewrite? An Undo pill appears for 8 seconds to put your original prompt back.
Click the ▾ next to Enhance for templates (style, lighting, composition, mood, quality) — selections clear after each run so they don’t pile up.
On phones the options open as a bottom sheet for easier tapping.

Prompt enhancement is live again

Click ✨ Enhance on the prompt box to rewrite your idea into a more detailed prompt.
Tuned per model — each image and video model gets its own phrasing tips.

Cleaner error messages

When a provider rejects a generation, the toast now shows the actual reason instead of the raw API response.
Your prompt and signed asset URLs stay out of error messages.
Applied across every image and video model.

OpenAI image editing

OpenAI gpt-image models can now edit existing images — drop a reference into Generate and they’ll follow your edit prompt directly.
Up to four reference images at once for compositing — describe how each should be used in your prompt.
gpt-image-1 and gpt-image-1.5 use high input fidelity automatically, tuned to preserve faces and fine detail across edits.
Defaults updated to follow OpenAI’s prompting guide — medium quality and opaque backgrounds for more consistent results.

OpenAI gpt-image-2

OpenAI’s newest image model is now available — sharper instruction-following and higher-fidelity image inputs.
Picked automatically when you select OpenAI in Generate; older versions (1.5, 1.0, Mini) live under the variant selector.
Heads up: gpt-image-2 doesn’t support transparent backgrounds yet — use 1.5 if you need an alpha channel.

In-app notifications

New bell in the sidebar shows submits, approvals, and rejections as they happen.
Unread count pill on the bell, with a “Mark all read” action in the popover.
Click a notification to jump straight to the brand’s review queue.
XP, level-ups, and badge unlocks now toast after every successful generation.

2026-04-28

Delete brands you no longer need

Brand managers can now delete a brand from the brands list (⋯ menu on each row) or from the brand’s settings page (new Danger Zone).
When deleting, choose what happens to the brand’s assets and brews: move them to another brand, or delete them along with the brand.
Type the brand’s name to confirm — same pattern as GitHub repo deletes.
Personal brands are hidden from this flow — they’re system-managed and stay tied to your account.
The brands list also stops showing other teammates’ Personal brands; you only see your own.

2026-04-27

Home page with quick actions

Renamed Overview to Home — same page, friendlier name.
New action strip up top: Text → Image, Image → Image, Text → Video, Animate.
Each tile drops you into Generate with the right mode preselected — Image → Image even picks Flux Kontext for you and opens the reference picker.

Move assets between brands

Asset detail panel now has a “Move to brand…” action so a miscategorized asset can be reassigned to the correct brand.
Available to the asset’s creator, brand managers on the source brand, and workspace admins — approved assets must still be forked, not moved.
Moving an asset resets its status to draft so the new brand’s reviewers can vet it.

Workspaces are now Studios

Renamed Workspace to Studio across the app — same thing, friendlier name.
New Studio settings page at /settings/studio for renaming, changing the slug, or pinning a logo URL.
The sidebar studio row now clicks straight through to Studio settings.

Campaigns

Group assets by initiative — create campaigns under any brand from /brands/[slug]/campaigns.
Brand managers can create, rename, or delete campaigns.
API for tagging assets with campaigns is live; gallery filter chip ships next.

Three-level brew visibility

Brews are now Private (just you), Brand (everyone on the brand), or Public (Explore tab).
Promoting a brew to Public still requires a brand manager — creators can flip Private↔Brand on their own.
Every visibility change is logged so you can see who shared what and when.

Sidebar reorganized — Overview, Personal, Review, then a brand list with one row per client/division.
New /overview page shows your drafts, your pending review queue, recently approved work, and personal stats.
Workspace switcher up top for users in multiple workspaces.
+ Add brand button for workspace admins; new brands land with a kit you can edit at /brands/[slug]/kit.
Per-brand pages live at /brands/[slug]/{gallery,brews,kit,members,review}.

Brand kit panel on the generate page

See exactly what your brand kit will inject before you submit — prefix, suffix, banned terms, default LoRAs, anchor refs.
Override toggle dims the panel and skips kit injection for one-off generations.
Personal brands skip the panel; their kit is empty by design.

Drag-and-drop uploads

New Upload button in the gallery — drop existing photos or short videos straight in.
Up to 50MB per file; supports PNG, JPEG, WebP, GIF, MP4, MOV, WebM.
Uploads land as drafts on the brand you pick, ready for the review pipeline.
Per-file progress with a cancel button — no surprise stuck uploads.

Gallery now scoped to your brands

Status badges on every tile — draft, in review, approved, rejected, archived.
Filter the gallery by status or brand; filters survive in the URL so you can deep-link a view.
You only see assets you created or that live on a brand you’re a member of.
Empty states distinguish “no matches” from “no access to this brand”.

Review queue (early access)

New Review tab in the sidebar with a pending-count badge for brand managers.
Submit a draft for review from the asset detail dialog — your brand manager picks it up from the queue.
Keyboard-driven approve/reject modal: j/k to walk the queue, a to approve, r to reject, n for a note.
Approved assets are now immutable — use Edit / Fork to start a new draft from an approved version.
Personal-brand assets stay out of the review pipeline as expected.

2026-04-25

OpenAI image models

Added gpt-image-1.5, gpt-image-1, and gpt-image-1-mini.
Native transparent PNG output for product and logo work.

New Help menu with Documentation, Report a bug, and About.
Sidebar now collapses to an icon rail (⌘B).
Reorganized navigation — account & admin moved to the bottom.

2026-04-24

Public brew pages with shareable links.
Browse and remix community brews from the Brews tab.

2026-04-22

HuggingFace LoRA support

Browse and load LoRAs directly from HuggingFace.
Improved LoRA browser with search and filtering.

2026-03-26

References

Every uploaded reference image is now tracked in a references library with full metadata: dimensions, file size, MIME type, thumbnail, and usage count.
New /references page accessible from the sidebar — a browsable gallery of all uploaded reference images with grid view, infinite scroll, lightbox detail view, download, delete, and Use actions.
The reference image section on the Generate page now offers two options: Upload (new file) and Browse (pick from existing). The Browse dialog has two tabs — Uploads (previously uploaded references) and Gallery (past generations to reuse as references).
Gallery lightbox now shows a Reference button on image assets, navigating to the Generate page with that image pre-loaded.
Gallery lightbox shows the reference image thumbnail in the parameters panel when a reference was used during generation.
Reference images (imageInput) are now saved and restored with Brews, so saved recipes preserve which reference image was used.
Usage count increments each time a reference image is used in a generation.

Added References entry to the application sidebar.

2026-03-25

LoRA support

Added full LoRA support for Flux image generation via the fal.ai flux-lora endpoint.
The system automatically routes to fal.ai when one or more LoRAs are active — no configuration change required.
Up to 5 LoRAs can be stacked per generation, each with an independent weight slider (0–4).
Trigger words are auto-inserted into the prompt when a LoRA is added.

/loras exploration page

New dedicated page for browsing the Civitai LoRA catalog.
Filter by base model: Flux (ready to generate), SDXL, Pony, Illustrious, SD 1.5, Hunyuan Video, Wan Video.
Sort by Most Downloaded, Highest Rated, or Newest.
NSFW toggle with preference saved locally per browser.
Detail modal with image/video gallery, trigger words, stats, and a link to the Civitai listing.
Video LoRA previews play inline as silent loops.

Brews

Brews are saved generation recipes that capture the model, LoRAs and weights, parameters, and optionally the prompt.
Save a brew from the Generate page after a successful generation using the Save as Brew button.
Load a brew on the Generate page from the Brew popover next to the model selector.
Each brew tracks a usage count that increments each time the brew is loaded.
Edit brew name/description or delete brews from the /brews management page.

Save brews from the Gallery

The asset lightbox in the Gallery now includes a Brew button.
Clicking it opens the Save as Brew dialog pre-filled with the model and parameters from that generation, allowing you to capture a recipe from any past result.

Added Brews and LoRAs entries to the application sidebar.

Changelog

2026-09-18

New pricing: plans by storage, free seats, credits for generation

2026-08-12

Upscale and Remove background

Upscale and Remove background from connected AI tools

Remove members, rename brands

Google’s image models, brought up to date

2026-08-09

Connect your AI tools to the studio

Studios can now subscribe

Workflow pages: shape the LoRA stack and tune every setting

Pick your studio’s look — color schemes and light mode

2026-08-08

Slack review threads now sync both ways

Folder import got a whole new home

Workflows are back — and now curated per studio

”Reject” is now “Request changes”

Studios are now invite-only

2026-08-07

A cleaner sidebar: studio switcher on top, account menu below

Approve reviews right from Slack

Connect your Slack account

Studio wins in Slack

2026-08-06

Slack notifications for asset reviews

Folder imports show live progress again

2026-07-03

Fixed image edits failing with a false “flagged as sensitive” error

2026-06-27

Browse your Library by date

Upload assets from anywhere with ⌘K

2026-06-26

We’re now called Cauldron

2026-06-24

Drop a whole folder into your Library and it organizes itself

2026-06-21

The Assistant can search your library and look things up

The Assistant can reframe your work and show it to you

Dark mode: muted text is easier to read

2026-06-20

Reframe, redesigned — a live preview and aspect ratios for any size

Choose the Assistant’s chat model — and the model that makes your images

A calmer, clearer Assistant

Rate the Assistant’s answers

2026-06-19

Reframe — crop a whole shoot for every social format at once

2026-06-12

Cleaner thumbnails

2026-06-06

Notifications you can actually scan — and act on

A faster ⌘K

2026-06-05

One consistent preview for everything you generate

Open any asset in place

Bring your own image — and @-mention — into the Assistant

Image batches that all show up

2026-06-04

Redesigned Home — search, or describe what to create

Voice typing

The studio fits your screen

A smarter Assistant

2026-06-03

Getting Started quest board

Redesigned Brands page with Gallery and List views

2026-05-31

Flux 2 Pro is the new default image model

Your studio’s name and logo in the sidebar

2026-05-30

Assistant moved into the main sidebar

Pick your brand once, everywhere

Quick actions in ⌘K search

2026-05-29

Library viewer is now a lightbox

Workflows now have dedicated pages

Tidier Library toolbar and reference search

2026-05-28

Composer redesign — model first, simpler settings

Generate up to 4 at once — results on the same page

Find similar, and search by meaning