Brand Studio

Brand Studio is one tool that turns a single Brand Kit into a full set of on-brand deliverables. It is built around Muse, an AI creative director who runs a bench of named specialists, working over the platform's shared image tools. Everything you make is matched to your own brand instead of drifting on every generation.

There are two ways to make any deliverable, and picking the right one is the whole trick:

1. Generate - the AI makes a NEW asset from a brief. Use this when you do not have the artwork yet.
2. Format my file - you already HAVE the artwork (a logo, a piece of art, a product shot) and you just need it correctly formatted for a platform. This starts from your exact file and applies only the steps the target needs. Use this when the asset exists and you want the right file, not a new design.

The Brand Kit

You can hold more than one. Use the caret beside the brand name in the header to switch, or add a new one. How many depends on your plan: Free includes one, Starter and Studio include five, and Artist and above are unlimited. Each kit carries its own name, colours, logos and reference images, and switching swaps the whole set - so one account can run several brands without them bleeding into each other.

Set up your Brand Kit once and everything feeds off it. It is saved to your account, so it follows you between sessions and devices.

• Brand name - the exact spelling is carried into every generation, so text on merch, posters, and covers comes out right.
• Reference images (up to 16) - your logos, album art, product shots, brand photos. They condition every generation so results reuse your real assets instead of inventing new ones. Uploaded images are hosted to your account.
• Reference priority (drag or Alt + arrow keys to reorder) - order is functional, not cosmetic. Reference 1 is weighed hardest and reaches every engine; later references only reach the higher-capacity models. Put your most important asset (usually your logo) first.
• Brand colors (up to 3) - click a swatch to set each color; they are applied across generations.
• Fonts (Title and Body, optional) - upload your brand fonts for your kit and to inform typographic advice. Note: AI image models cannot bake an arbitrary uploaded font into generated text, so exact typography is a separate step; the fonts are stored for reference, not baked into the pixels.

Open the Brand Kit from the brand name in the header or the "Review kit" button. Add, replace, remove, and reorder references there.

Generate: make a new asset

Pick a discipline, then fill in the brief and click the pink Commission button:

• Format - the placement or target (for example a merch placement, or a social target like avatar or story).
• Fidelity - Fast (a quick draft, Grok Imagine 2.0), Best (sharpest text and logos, GPT Image 2.5), or Print 4K (true 4K, Nano Banana 2 HD). If a format's shape is one an engine cannot make, it renders the closest shape it can; Format my file crops to the exact size afterwards.
• Direction - a free-text style, with one-tap suggestion chips you can stack. Leave it blank to let your references drive the look.
• Description - what you want, in your words.
• Count (1 to 4) - how many options to generate.

Your Brand Kit conditions the result: the name, the colors, and (when present) your reference images are passed in so the output looks like you. Results appear in the deliverables grid; the first is marked the Keeper. Generation uses credits and requires you to be signed in. Finished images download straight to your device.

Format my file: correctly format artwork you already have

When a discipline supports it, a Format panel appears. Pick a Brand Kit reference or upload a one-off file, then click the Format button. It starts from your exact file and applies ONLY the steps the chosen spec needs, in this order:

1. Reframe - if your file's aspect ratio differs from the target, it is content-aware reframed to fit (so your subject is preserved, not cropped away).
2. Upscale - if the target is bigger than your source, it is genuinely upscaled to add resolution.
3. Remove background - if the spec needs transparency (merch, stickers, emotes).
4. Finish - a precise resize or crop, a multi-size set, or a real 300-DPI print stamp.

A plain downscale (a large photo into a smaller post) is instant. Steps that use a model (reframe, upscale, remove background) use credits, so the cost tracks the work. The result tells you which steps were applied. Files download to your device.

The disciplines and what they produce

UGC Ad and Edit are generation and editing flows, not simple file formats.

UGC Ad: a whole ad in one pass

Pick a spokesperson, add your product, and choose Photo or Video at the top of the brief.

Photo makes a still ad on GPT Image 2.5 Sunburst: your spokesperson holding your product. Fidelity picks the resolution - Fast 1K (9 credits), Best 2K (14), Print 4K (23).

Video makes a talking ad with native audio. The engine films the scene and speaks your words in the same pass, so the voice and the mouth are made together. There is no separate voice step, no lip-sync step and nothing extra to pay for the audio. It is the same method UGC Studio uses.

The brief, top to bottom:

• Scene & movement - where they are and what they do. Describe the room and the action, not their face; your spokesperson image carries the face.
• What they say - your voiceover. It is spoken word for word, in English, and nothing else is said. Leave it empty for a wordless clip with natural room sound. If this box is empty, any lines you put in "quotes" inside the scene are spoken instead.
• Voice - a written description of how they sound: age, accent, pitch, pace, energy. Pick one of ten stock voices (Amy, Jessica, Laura, Katie, Sarah, Liam, Eric, Roger, Brian and Bill) and press Sample to hear a real render of it, or write your own. Name and Save a description to reuse it: the same words every time is how a campaign keeps one voice. It describes a voice; it does not clone one, and it cannot use a recording you already have.
• Engine - five chips, each showing its live price per second:

• Length, Shape (vertical, square, wide) and Quality (480p or 720p).

The take is as long as your lines need. Speech runs about 2.3 words a second, so the take is lengthened to fit what you wrote (a 20-word line needs about 10 seconds), and the price under the button already includes it. If your lines need more than the engine's longest take, the brief tells you: trim the script, or switch to 2.5 or Grok for up to 30 seconds.

The rate is the same whether the clip speaks or not - across the Seedance 2.x engines the price with audio equals the price silent. A clip that fails is refunded automatically, and one that is still rendering when the page stops waiting lands in your gallery by itself, so do not run it again.

Want the same faces posting for weeks? UGC Studio is the dedicated app for a recurring team of characters, each with a locked face, personality and voice. Brand Studio's UGC Ad is the quick, single-ad version of the same method.

Edit: one board for changing an image

Anything that already exists goes here - send it with the Edit button on any result card, or upload one. The board is deliberately large so a tight selection is workable. Six tools:

• AI Edit - describe the change in words. Nano Banana 2 Lite (a flat 6) or Grok Imagine (a flat 10). Anything you do not mention is meant to survive untouched.
• Clean edges - one click to smooth jagged or pixelated edges, a flat 6. Run it before Cut out.
• Extract - drag a box on the board and take that piece out. Free - the crop happens in your browser at the image's full resolution, so nothing is generated and no credits are spent.
• Upscale, Reframe, Cut out - the model steps, which do use credits.

Every tool acts on the image currently on the board, so upscale, then cut out, then extract is one sequence rather than three trips through your downloads folder. Undo and Redo step through the history. Upscale BEFORE you cut out - doing it after just enlarges a soft alpha edge.

The Logo Pack

If you already have your logo, the Formatter's Logo pack target formats your EXACT logo (a Brand Kit reference or an upload) into a ready-to-use set. It does not regenerate your logo, change its text, or alter it in any way.

The pack downloads as a single zip, and everything in it is built from a true vector master.

One lockup per run. The pack formats the single logo you selected. A full identity is usually three lockups - primary, horizontal, and symbol-only - and each is its own pack run, so a complete set is three packs. Lockups are different arrangements of your mark rather than different file formats, so they cannot be produced automatically from one source.

The vector files (the important ones)

• -logo.svg - the full colour master. A vector is maths rather than pixels, so it stays sharp at any size: a business card and a building sign come off the same file. Send this one to printers, sign makers and merchandisers. It opens in Illustrator, Affinity, Figma and Inkscape (free).
• -logo-black.svg and -logo-white.svg - one-colour versions. Embroidery, engraving, laser cutting, foil, rubber stamps, screen printing and vinyl usually cannot reproduce colour or gradients, so they need a single flat shape. Black is for light surfaces, white for dark ones. Most vendors will ask for one of these, and a pack without them stalls the job.

The raster files

• -LOGO.png at 2048 px, plus -logo-4096 / 2048 / 1024 / 512.png - all transparent. Use the size closest to how it will be shown; handing a browser a 4096 px file to display at 200 px wastes bandwidth and can look softer than the 512 would have.
• -logo-black-2048.png and -logo-white-2048.png - the one-colour versions as raster.
• White-background and black-background PNGs - flattened previews for checking contrast and for places that will not accept transparency. They are previews, not masters.
• Favicons at 512, 192, 180, 32 and 16 px - the app-icon sizes (180 / 192 / 512) carry a safe-zone margin so the OS does not clip your mark when it masks icons into rounded squares or circles. The tab favicons (16 / 32) are deliberately tight, because at that size a margin makes the mark unreadable.
• A print PNG at 4096 px stamped 300 DPI. DPI on its own means nothing - it is only a ratio between pixels and a physical size. That file is 13.6 inches at 300 DPI and 27 inches at 150 DPI, which is what direct-to-garment printing actually uses. If a printer needs a size it cannot cover, send the SVG instead; it has no limit.
• README.txt - explains what every file is for, so a client who did not build the pack still knows what to do with it.

What is not included

.eps and .ai are not in the pack. Most printers accept SVG or PDF; if yours insists on EPS, open the SVG in Illustrator or Inkscape and export one.

Stacked, horizontal and icon-only lockups are not included either. Those are design decisions about your mark rather than file conversions - open the SVG and rearrange it.

How to build it: open the Formatter in the left rail, choose your logo (upload it, pick a Brand Kit logo slot, or send one over from a result), pick the Logo pack target and click Format. "Remove the background first" is on by default, and it is skipped automatically (and not charged) when your logo is already transparent - so you can normally leave it alone. The zip downloads to your device.

Muse's Network

Muse is your in-app creative director. She answers questions, has real taste, and reaches seven subject matter experts for precise production facts: Logo Station, Merch Station, Print Lab, Social Desk, Type Foundry, Brand Family, UGC Desk (scripts, voice descriptions and talking-ad craft). The panel shows the expert whose specialty the discipline you are in happens to be, plus whoever she consulted last; the rest are one click away. A persona dial (Just the facts / Collaborator / Companion) sets her tone.

Muse gives advice and guidance - she is not able to click buttons or run anything for you. She will tell you exactly which control to use (Commission to generate, the Formatter to format a file or build the Logo pack), and you perform the action. She will never claim to have built, formatted, or delivered something herself.

She can see your work

Muse looks at whatever is on your board. When results are on screen she sees those, up to four at once; when the board is empty but you have an image loaded on the Edit station, she sees that instead. So you can ask "which of these is strongest?" or "what would a client push back on?" and get an answer about your actual artwork rather than a general opinion.

A Critique button appears above the message box whenever there is something for her to look at, and it is hidden when there is not. That is deliberate. If she cannot see the work, she should not be reviewing it.

Two things worth knowing:

• She reads the finished image, not your brief. If a detail matters and is not visible in the picture, tell her.
• A turn where she is looking at images takes a little longer than a plain chat reply.

Notes

• Brand Studio uses the platform's shared image tools, so generation and any model step (reframe, upscale, remove background) use credits and require you to be signed in.
• A plain resize or crop in Format my file is free and instant; only model steps cost credits.
• All results download straight to your device. Multiple pack files download in sequence, so allow multiple downloads if your browser asks.
• For the most on-brand results, fill in your name, colors, and a few references (logo first) before you generate.