Just say what to show. Kinisma opens your app, clicks through it, records the real screen, then writes and voices the narration — fully automatic, about $0.60 a minute.
Name the app and the task. The agent drives a real browser, films every step and narrates each click. A 1-minute tutorial costs about $0.15 in AI; longer ones and revisions cost more.
“A tutorial on Carbon (https://carbon.now.sh): how to turn your code into a beautiful image. Paste your code, pick a theme, change the background colour, and export a PNG.”
It reads your site and help pages, then walks the flow in a real browser to learn it.
It opens a browser window for you to sign in. The agent never sees or types your password, and never creates accounts.
Click by click in full HD, each recording carrying on where the last one ended. Most personal details on screen, like emails and phone numbers, are hidden.
Chapters with title cards, and one spoken sentence per step, with a zoom and a callout on every click.
A scene writer per scene, all in parallel, writes real motion graphics (sometimes 3D), compiles it and looks at stills of its own scene.
Text that’s too long, timing against your chosen length, then a director’s review of every scene’s stills, fixing what’s clearly wrong.
A sketch, title, length and narration for every scene. It stops here until you approve, or ask for changes in plain words.
The voice you picked reads the script, then speech-to-text times each word so the visuals land on it. Each click lands on its sentence.
An original track composed by AI (about 8¢), or a free Creative Commons track with its credit, mixed under the voice.
It plays right away in the app. Say what to change, for the whole video or one scene; only what changed is rebuilt.
Export renders a 1080p MP4. Publish writes your titles, descriptions with chapters, hashtags and three thumbnails, ready to paste.
Real videos built by the Video Agent. Press play to watch.
The agent uses your real app and films every step. You don’t capture anything.
One sentence per step, timed so each click lands on its words.
A zoom and a callout on every click, with chapter cards between sections.
You sign in yourself in a browser it opens; it never sees your password.
Emails, phone numbers and account names are hidden by default. Check names in tables.
Publish writes a description with chapters from the real scene times.
“Paste your code, pick a theme, change the background, export a PNG” becomes the chapters.
Start it on the right page and it finds the path faster.
Sign in with test data, so nothing private shows on screen.
Supademo, Arcade, Guidde, Scribe and Loom are great at what they do — but they all start with you recording each step. Kinisma starts with one sentence.