What an AI video agent is, and what ours decides
Software you give a goal to. It plans the video and waits.
Already have an account? Sign in
Free to start · No credit card
What an AI video agent is, and what ours decides
An agent takes a goal, not a prompt. It plans the whole video as scenes, prices the plan, and waits for you to say yes. Here is what that means in practice, and where ours draws the line.
Idea to video, with a stop in the middle where you read the bill.
Ours is called Kae. She plans on the storyboard. Approving, rendering and editing stay in your hands.
Part of AI video generator.
What is an AI video agent?
An AI video agent is software you give a goal to, not a prompt. It reads the goal, plans the whole video as scenes, picks the presenter, the script, the shots and the voice, prices the plan, and waits for you to approve it. Ours is called Kae. She plans. She never renders anything on her own.
Who it is for
Someone who knows roughly what video they want and would rather not decide fifty small things to get it.
What you need
A sentence about the video and a Workspace. A Digital Twin, a saved voice or footage you already filmed will make the plan better. None of them are needed to start.
What it costs
Planning sessions are included in every plan: 5 a month on Free, 20 on Plus, 60 on Pro, and no monthly count on Max or Ultra. The renders you approve cost credits. Talking-head video is 152 credits per second on the paid default setting, and a Free Workspace starts with 8,400 credits, once.
What it does not do
It never renders without your approval, and it does not edit for you.
Agent, generator, assistant: the plain difference
The three words get used for the same thing. They are not the same thing.
A generator takes a prompt for one clip and hands back one clip. You decide everything, shot by shot.
An assistant takes a question and hands back ideas, words, a draft script. You decide everything that comes after.
An agent takes a goal for the whole video and hands back a plan for every scene, priced. It proposes, you approve.
A generator is a good tool when you already know the shot. An assistant is a good tool when you are stuck on the words. An agent is the one that holds the whole thing at once, which is why the approval step matters so much: it is the only place a person is still in the loop.
The four things that make it an agent
Four things separate an agent from a text box that makes clips.
- 1
It takes a goal.
You say what the video is for and who it is for. You do not write a prompt for every shot.
- 2
It plans in steps.
The goal becomes scenes, in order, each with the words that get spoken in it.
- 3
It chooses the pieces.
Every scene gets a kind: your face speaking, a cutaway, a still that starts moving, a clip from a library. It picks who presents each scene and which voice reads the lines.
- 4
It shows the bill first.
The plan carries a cost on every line and a total at the bottom, before anything runs. Skip this one and you are approving a bill you have not seen.
What ours decides, and what it hands back to you
Kae decides.
The order of the scenes. The words spoken in each one. The framing and the camera move. Who presents each scene, from the characters already in your Workspace. Which voice reads it. Where a cutaway belongs and what is in it. Whether a scene starts from a still and then moves, or goes straight to video.
Kae hands back to you.
The quality level, which you pick. The approval, which is a button you press. The money, which only moves once you have pressed it. The edit, which is yours from the moment the scenes exist. One rule in there is worth saying out loud: a plan whose quality level has not been chosen is not priced at all. She asks the question instead of picking a default and quoting you on it.
A brief, planned: what the storyboard actually holds
Say this to her: a forty-five second explainer for our new service, in my voice, with two product shots and captions.
What comes back is a storyboard, not a video. Four scenes in order. Three of them are you speaking, with the exact words written out, a framing and a camera move on each. One is a cutaway made from a sentence. Your voice sits on all four. Every scene has its own line on the plan card with its own cost, and the card carries the total. Nothing has run.
Then you change things, because the first plan is rarely the last one. Rewrite a line. Swap the presenter. Drop the third scene. The plan updates and the cost updates with it, and you read the new total before you approve.
The numbers behind those lines come from one place. Talking-head video is 152 credits per second on the paid default setting. Freeform cutaways run four to fifteen seconds on the main settings, on every plan. A Free Workspace starts with 8,400 credits, once, and a free render runs up to one minute. The card does the arithmetic so you are not doing it in your head.
When the scenes are made, the editor opens with them already laid out. Cut, layer music, burn in captions, export. Exports cost 20 credits a second of finished timeline, and never less than 60 credits.
Six questions to ask any AI video agent
Take these to anyone, including us. They are the questions that separate a planner from a spender.
1. Does it show the cost before it runs, per item and as a total? Ours does, on the plan card, worked out again on the server every time the plan changes.
2. Does anything start without you pressing approve? No. Approving is a button a person presses, and approving by itself makes nothing. The scenes go out one by one through the same doors you would use by hand.
3. Whose face and voice can it use, and who wrote the permission down? Only a Digital Twin and the voices already set up in your Workspace, each with a consent record behind it. She will not build a person from a photo of a stranger.
4. Can you change what it planned, or only ask for another one? Change it. The storyboard is editable scene by scene, and the price follows your edits.
5. When a scene fails, what happens to the money? It comes back. A scene that fails refunds through the same machinery that charged it, because there is only one place in the product where credits move.
6. What does it hand you at the end: a video file you download, or a link inside their tool? A file. You download it.
If a tool cannot answer the first question or the second plainly, it is not planning for you. It is spending for you.
Consent and disclosure, since the person on screen is real
An agent that plans a video of a real person is planning to use someone's likeness, so the permission has to come first. Kae can only put a Digital Twin on screen when that Workspace holds a current consent record for it, and she will not build one from a photo of someone who did not agree. An Avatar is the other kind of character, a made-up one with no real person behind it, and it needs no consent because nobody is being used.
On the output side, machine-verifiable provenance is being rolled out and is not yet guaranteed on every Render. Renders made on a Free plan carry a visible mark. Telling your audience that a video was made with AI is your call, and it is your duty wherever the rules in your market ask for it. Read how we handle consent, disclosure and content credentials.
What it will not do
Three limits, said plainly, because you will find them anyway.
She never renders on her own.
There is no setting that lets her. Approve is a button a person presses, and even then the renders go out through the ordinary billed doors rather than some second path of her own.
She plans on the storyboard, not in the editor.
Once the scenes exist, the editor is a manual room. You cut, layer, caption and export yourself. She does not sit in there rearranging your timeline.
She does not publish or schedule anything.
Kyndrify makes the files and hands them to you. Where they go next is your decision and your calendar.
What people ask about AI video agents
- What is an AI video agent, in one sentence?
- It is software that takes the goal of a whole video, plans every scene of it, prices the plan, and waits for a person to approve before anything is made.
- How is that different from typing a prompt into a video generator?
- A generator answers one prompt with one clip, and you do the joining up. An agent holds the whole video at once: the order of the scenes, the script, who presents, which voice, and what each part will cost.
- Does it make the video by itself?
- No. Kae plans it. When you approve the plan, the Studio renders each scene through the same tools you could use by hand. Then the editor is yours.
- What does planning cost?
- Planning sessions are included in every plan: 5 a month on Free, 20 on Plus, 60 on Pro, and no monthly count on Max or Ultra. Only the renders you approve spend credits.
- Can I change what it planned?
- All of it. The storyboard is editable scene by scene, the cost updates as you change it, and you approve before anything is made. You can also leave a comment on a single shot and share the link with a colleague.
- Can it use my own face and voice?
- It can use the Digital Twin and the voices set up and consented to in your Workspace, and nothing else. It never invents a person and never uses someone else's likeness.
Also worth a look
More on this from the blog
See a plan for your idea
Say what the video is for. Kae writes the plan, prices it, and stops there until you say yes. Free plans include 5 planning sessions a month, and no card is needed to start.