Images, video, documents and sites
Your employee makes what goes to clients and into presentations by itself: images, videos, documents, pages and sites. You ask in words, as you would a designer; there is nothing to switch on.
What images does it make?
Lifelike scenes with people and products, banners, product cards, diagrams, covers and posters. Text on an image, a sign or a price tag, it writes word for word and exactly where you asked. Send a finished image and say what to change: exactly what you named changes, and the rest of the frame stays. A frame takes under a minute. Proportions are set in words: a vertical 9:16 frame, a 1:1 square. A logo or an icon it can also draw as a vector SVG file, and for an exact size, a transparent background or replacing an object in a photo it picks a suitable model itself.
And video?
An announcement, an ad, an explainer, an animated chart or a reel, with music and sound, in 16:9, 9:16 and 1:1. Describe the video as you would to a designer: the employee makes it and checks every frame and the loudness itself. Send your own raw recording, and it edits it: cuts pauses and slips, adds graphics on top. The style can be anything, even anime or comic. It helps to have your logo as a file, your colors, fonts and real shots of the product. The voice in the video is its own by default; on request it picks another one to suit the brand.
What does the employee itself look like?
It has a portrait and an animated version, and on request it changes its avatar. Send your own picture and it uses it as is; with no picture, it draws a portrait to match its character and changes the hair, clothes or background if you ask.
Which documents does it work with?
- Word: a contract, a statement of work, a commercial proposal, a report; in your file it can make changes in track changes mode and leave comments.
- PDF: read, pull out text and tables, merge, split, fill in a form, recognize a scan, build a new one.
- Excel: calculate, fix, build a chart, clean up an export; formulas stay formulas, and your template's formatting is kept.
- PowerPoint: a pitch deck, slide edits, work with a template and speaker notes, your brand colors.
Will it lay out and publish a page?
Yes. Give it a picture of the page drawn by any model, or a screenshot: the employee lays it out, opens it in a browser, compares it with the picture and fixes the differences. It publishes the finished site at its own address on emagi.ai, at a separate address for the project or on your domain; for a domain, the employee tells you which record to set in the domain settings. What is published is open to everyone without a password, so only the site itself goes out, without working files. What is published is shown on the agent's page in the "Workspace" tile.
What uses media credits and how do you buy more?
Images, video, voice messages, transcription and call minutes. Images drawn on your own ChatGPT subscription do not use credits. Your plan includes a monthly media credit allowance, five times larger on VIP. How much of the monthly allowance is used is shown on the "Media credits" card on the dashboard home page. When credits run out, generation waits for the next period and your bill does not grow; a call will not start then, and the employee says why.
If you need more, press "Buy more credits" there: before paying you see how many credits a month the pack adds and what it costs. Bought credits go on top of the allowance. If you want to approve spending yourself, say so: the employee will name the price before starting.
Read next: Your AI subscription and models · Dashboard · Calls: your employee talks by voice