Making a video

Two different things are called "making video", and they are not in the same screen.

What you want	Where
Have one clip generated, a few seconds long	Single video, in the Creative Studio
Put images, audio and text on a timeline and export	Video Studio

A generated clip is source material. The edit is what goes out. You often need the second and sometimes the first.

On credits or with your own account

Without an account of your own you make a clip on Denkploeg credits. You pick a mode, not a model:

Mode	Model	Length	Roughly, cost per second
Fast	Veo 3.1 Lite (Google), 720p	4, 6 or 8 seconds	0.05 dollar
Standard	Veo 3.1 Fast (Google), 720p	4, 6 or 8 seconds	0.10 dollar
Expert	Veo 3.1 (Google), 1080p	8 seconds	0.40 dollar

If Google is briefly unavailable, Kling 3 (fal) makes the clip instead. The price in credits is on the button before you make it, and a clip that fails costs nothing. Clips have sound. Product photos as references (up to three, from your library with known rights) work on Standard and Expert: the clip is then always 8 seconds and landscape (16:9), without a start image, and there is no fallback to Kling. It is not available on Fast. The price is on the button beforehand.

A UGC video on credits works per scene: the quote is in Denkploeg credits, and that is the amount you sign.

Single video: three modes

The mode determines what you send along, and the provider knows three of them.

1. From text

Only a description. The model invents everything: the person, the place, the light. Quick for seeing an idea, and useless as soon as it is about your product, because the product does not exist there.

2. From a start image

You send one image along and the clip starts there. This is almost always the mode you want: you have already judged the image, so you know what it looks like, and the generation only has to invent motion.

Some models also accept an end frame. Then the model knows where it is heading instead of picking a direction itself. If the chosen model does not know that field, it is quietly dropped and the screen says so.

3. From references

Multiple images: a face, a product, an environment. This is the mode for a clip in which your model wears your jacket.

Note: reference space is tighter for video than for images. Four views of a face plus a product photo is already a lot. If an anchor is ready, it goes first and pushes out the loose model images, and that is exactly the intention: one approved image works better than four that differ from each other.

Camera motion is text, not a button

With this provider, camera motion is plain prompt text. There is no parameter for it. That is why there is a text field and not a dropdown: a dropdown would suggest that we steer something that we do not steer.

Write it as a camera direction:

slow push-in on the jacket, handheld, slight sway

static wide shot, the model walks in from the left and stops in the centre

What does not work: "dynamic", "cinematic", "professional movement". Those are judgements, not movements.

Steps that save the most credits

Make the image first that becomes the first frame, in the Creative Studio. A clip is more expensive than an image, and a clip that starts from an image you have already approved rarely fails on the subject.
Choose the shortest duration that shows the idea. The price scales with the length, and a four-second clip shows whether the motion works just as well as an eight-second one.
One movement per clip. A camera that zooms in while the model walks and the background changes produces three half things.
Watch it back before you extend it. The second second is where faces drift away.

The screen tells you in advance what the clip will cost and asks for confirmation. If it does not know the price, it refuses.

The Video Studio: from clips to an edit

The Video Studio is not a generator. It is an editor with a timeline, and what is on it comes from the library of the same workspace: generated clips, supplied images, audio.

The whole screen is described in Video Studio. What you need to know to work with it smoothly:

An edit is a document with a history. You close it, you open it a week later, and it is exactly what you left. Every change is a revision; nothing is overwritten on the spot.
The library is shared with the Creative Studio. An image that was approved there is ready here. Putting it on the timeline is up to you: nothing happens automatically with your edit.
Text comes from us, not from the model. The text layers use the typefaces and colours from the brand profile. That is the same rule as for images, and for the same reason: a second language then means setting the text again instead of generating again.
Spoken text can become subtitles. That runs on a transcription of the audio that is already there; it does not invent words that were not spoken.
Rendering happens in its own process. You can close the tab. The queue survives that, and a failed render says what went wrong instead of twenty lines of ffmpeg.

What goes wrong, and what helps

What you see	What is going on	What helps
The face drifts away halfway	No anchor, or too many different references	Set an anchor and leave out the loose model images
The clip does something completely different from what was asked	Motion described as a judgement	Write a camera direction, not a mood word
The product changes shape	No start image	Generate the image first, use it as the first frame
The model does not know the end frame field	Not every model has it	The screen says which fields it does accept
The export looks different from the preview	Different scale or encoding	Render a short one first and check it before you make the long one

What this does not do

No automatic edit from a brief. There is a step that makes an edit proposal, but that is a proposal you open and adjust, not an export that is ready by itself.
No voice cloning and no lip sync.
No publishing to an advertising platform. Delivering is a file plus the file around it.

Via MCP and the director

An agent can read and save UGC projects (creative_list_ugc_projects, creative_get_ugc_project, creative_save_ugc_project), estimate the cost (creative_estimate_ugc_project, creative_estimate_image, creative_estimate_video) and start a project (creative_start_ugc_project). Starting costs credits and carries the approval of the estimate (revision, time and amount): the agent should show you the amount before it starts, and the server refuses an amount nobody saw. The emergency stop halts it. Names and texts from a project appear in the answer under onvertrouwd.

A proposal that a human signs: creative_ugc_archive proposes archiving a UGC project: it leaves the list, and plan, revisions and clips stay stored. Nothing changes until someone signs under Approvals. The agent asks for the outcome with creative_voorstel_stand.
Making an empty video (video_maken) and deleting a video (video_verwijderen) are described under Video Studio.

Navigate