MCP connector · ChatGPT

ChatGPT product photos for fashion brands
Your catalogue in. Your frames out.

ChatGPT can already make a picture. It cannot make your bag, in your light, on the face you cast last month. Connected to Misu, it can ask a studio that knows all three.

Beauty portrait of a model with a slicked-back ponytail holding a black dropper bottle labelled Misu Frame Serum beside her chin, against a warm beige backdrop.

01

What changes when Misu is connected

A general image tool works from words. Describe a jacket and you get a jacket, not yours. For a product page that is the whole problem: the stitching, the hardware and the label have to be the ones on the shelf.

Misu is a creative studio for fashion and product brands, and it holds what a chat does not: your products with their reference photos, the characters you trained, your saved Looks and your credit balance. Its connector is an MCP server. Add it to ChatGPT and the conversation can read that catalogue and ask Misu to render from it.

ChatGPT does the talking. It plans the frames, asks the questions a producer would ask and reports back. Misu does the photography, on the same engine as the Studio in the browser, with the same product references, the same Character Lock and the same credit ledger.

02

Add Misu to ChatGPT

This runs through ChatGPT's developer mode on the web, on Plus and above. Misu is added as an MCP server by its URL. There is no API key to manage.

  1. 01

    Copy the connector URL

    In Misu, open Settings, then Connections, and copy the URL. The same URL is on the MCP page of this site.

  2. 02

    Switch on Developer mode

    In ChatGPT, open Settings, then Apps, then Advanced, and switch on Developer mode.

  3. 03

    Add an MCP server

    Add a server and paste the Misu URL. ChatGPT registers itself with Misu on its own; you do not create a client or a secret.

  4. 04

    Sign in and approve

    You are sent to Misu to sign in with your existing account. The consent screen names what the connection can do. Approve, and you land back in ChatGPT.

  5. 05

    Ask which workflows there are

    ChatGPT does not show a prompt menu for connected servers, so ask it: which Misu workflows are there? It lists the ready-made recipes and starts the one you pick.

03

What ChatGPT can do with your brand

Thirty-three tools sit behind the one URL. Reading is free. Rendering spends the brand's credits.

See what you sell

Your brands, your products and the characters and Looks you have saved. ChatGPT works from ids, so a request points at the real product.

See the price first

It can list the models Misu offers with their prices and typical render times, and quote a whole job in credits before anything runs.

Shoot stills and sets

A single frame, or up to eight different frames in one batch. A failed frame never stops the others, and only that frame is retried.

Turn a still into a clip

Video is priced with a dry run, submitted on your yes, then polled. Clips take one to four minutes. That is normal.

Turn a product photo into 3D

One product photo becomes a 3D model, returned as a GLB file and as STL for printing.

Find what it made

Everything lands in the brand's Library, tagged as agent work, and ChatGPT can browse recent work to pick up where the last chat stopped.

04

Prompts a brand would actually type

Each one is a real workflow that ships with the connector. In ChatGPT you ask for it in plain words.

"Use Misu to make product page photos for our leather tote: front and three-quarter on white, one detail of the clasp, one lifestyle frame."

ChatGPT looks the tote up in your catalogue and asks which one if several match. It quotes four images, waits for your yes, renders them in one batch, and hands back a table of frame, status and link with the credits spent.

"Show our bestseller in five colourways, same angle and light in every frame."

It asks for the colour list if you did not give one, writes a single base prompt and changes only the colour per row. You get five links in your order.

"Here's a photo of an outfit. Turn it into a three-frame campaign."

ChatGPT cannot pass a chat attachment to Misu itself. It gives you an upload link instead; you drop the photo in, and it picks the file up from your Library. Then it agrees a setting and a mood, quotes, and renders three frames with the outfit as the reference.

"What would a six-second clip of that last frame cost?"

A free answer. ChatGPT names the model, runs the dry run and gives you the credits. Nothing is submitted until you say so.

05

Credits, permissions and what stays in your hands

Free: reading brands, products, characters and Looks, checking the balance, listing models, quoting a job, checking a render and browsing the Library. Paid: stills, batches, clips and 3D models, at the same rates as the Studio. An image is typically 4 to 120 credits depending on model and quality. A failed render refunds itself.

The connection is your own Misu sign-in, not a shared key. ChatGPT reaches the brands your account can reach. The server resolves a brand from its name and checks your membership every time; it never trusts an id the assistant made up. A teammate who cannot spend in the Studio cannot spend from ChatGPT either.

Two limits protect the balance. A batch that would cost more than 1,500 credits is refused before anything is charged, so a larger set is split. A single video call is refused above 3,000 credits unless you raise the limit yourself.

06

MCP or the Studio

The chat is a way to brief. The Studio is a place to work. Both draw on the same credits and fill the same Library.

How you ask
In ChatGPTA sentence
In the StudioControls and panels
Product accuracy
In ChatGPTProduct ids and locks, as in the Studio
In the StudioProduct ids and locks
Frames per request
In ChatGPTUp to eight in a batch
In the StudioSet by the tool you are in
Ready-made workflows
In ChatGPTAsk for them by name
In the StudioCampaigns, Canvas, Ads
Uploading a photo
In ChatGPTThrough an upload link
In the StudioDrag and drop
Retouching, try-on, training
In ChatGPTNo
In the StudioYes
Cost
In ChatGPTStudio rates, quoted first
In the StudioStudio rates

07

Questions people ask

Is Misu an app in the ChatGPT app directory?

No. You add it yourself as an MCP server by URL, with Developer mode switched on. It takes four steps and no API key.

Which ChatGPT plan do I need?

Developer mode is on ChatGPT for the web, Plus and above. That is the route the steps on this page follow.

Why not use ChatGPT's own image generation?

For a mood image, do. For a product you sell, the picture has to show that product. Misu renders from your catalogue's reference photos and trained locks, and keeps one face across a campaign.

Can I attach a photo in ChatGPT and have Misu use it?

Not directly. ChatGPT asks Misu for an upload link, you drop the file there, and it is then available to render from. A photo that is already at a public URL needs no upload.

How do I start a workflow in ChatGPT?

Ask which Misu workflows there are. ChatGPT lists them: packshot set, colourways, Character Lock campaign, outfit to campaign, still to motion and a brand check. Then name the one you want.

Does ChatGPT show the images in the conversation?

Yes, as links and small previews. Where the chat app supports it, results open in a Misu panel with a gallery, a player for clips and live status for anything still rendering.

What does a ChatGPT photoshoot cost?

The same as in the Studio, paid from the brand's credits. Ask for a quote first; it is free. New studios start with welcome credits.

Can my whole team connect?

Each person connects with their own Misu account. Their chat can do what they can do in the Studio, so a member without permission to spend cannot render.

Get started

Your product.
The same face in every shot.

Full commercial rights · Licensed models · Stockholm

Misu is in beta and open to brands, agencies, freelancers and designers. We review every request and email your invite once it's approved, and every new studio starts with welcome credits.

BETA · REVIEWED BY HAND · FOUNDING PRICING LOCKED