Media Studio

#What Media Studio does

Describe an image and get one that looks like your brand — and write a script to get a voiceover read in one of forty voices.

• Four choices, four jobs. Sharp graphics and logos; photorealism; highest fidelity; or quick cheap drafts. The first two are the same model steered differently — you pick by what you are making, not by model name.
• The prompt is improved first. What you type is not what reaches the model. Refinement happens before generation, so a short description still produces a considered image.
• Scored, not just returned. A vision model rates the result and retries once if it misses. You see the good attempt, not the first one.
• Straight into your brand library. Save a result to Brand assets in one action, so a good image becomes part of the library instead of a file in someone's downloads.
• Voiceovers too. Write a script, pick from forty voices with samples, and get an audio file saved in your own storage — with its own library beside the images.

#Why it exists

Generic AI images look generic. You prompt, you get something almost right, you prompt again — and the result still does not match the brand you spent months defining.

Your prompt is refined before it reaches the model, and a vision model scores what comes back and retries once if it falls short. Four choices across three models, each tuned for a different job.

#Generate an image

1. Describe what you want. Your prompt is refined before it reaches the model, so a short description still produces a considered image.
2. Pick a model by the job: sharp graphics and logos, photorealism, highest fidelity, or a quick cheap draft.
3. A vision model scores the result and retries once if it falls short — you see the good attempt, not the first one.
4. Save to brand assets so a good image joins your library instead of a file in someone’s downloads.

#Generate a voiceover

The studio does audio as well as images — useful when a campaign needs a read, not just a picture.

1. Write the script.
2. Pick a voice — there are forty, each with a sample so you choose by ear rather than by name.
3. Generate. The audio file is saved in your own storage, in its own library beside the images.

#On the MCP surface

Your own AI can generate images over MCP, grounded in your brand definitions — so a social post can be drafted and illustrated in the same conversation.

#Works with

• Brand Assets. Logos, imagery and icons — tagged and findable.
• Social. Connect your channels, publish and schedule.
• Design System. Your brand as structured design tokens.