10% off monthly plans with code NODE10, until August 31
Your product filmed like an ad: a Camera node at 85 mm and f/1.4, a slow dolly in, and eight seconds of 1080p video with sound. The full workflow, no account needed.
Real workflow, real results. Pan and zoom freely.
Use this workflowThe real workflow: the Camera node settings flow into both the image and the video, the bottle is lit in close-up then filmed on a slow dolly in.
At a glance
The workflow is four nodes. A product card, meaning a precise description plus two or three photos, says what to show. A Camera node says how to show it. Both feed an Image node where Seedream 5 Pro composes the scene, then a Video node animates the shot.
The Camera node is what separates an AI picture from an actual shot. An 85 mm lens at f/1.4 crushes the background and lifts the product off it, a low angle makes it imposing, a slow dolly in makes the camera feel like it is closing in. Those settings are wired into both the image and the video: the framing you describe becomes the framing you get, and nothing is lost between the two steps.
The video comes out at eight seconds, 16:9, 1080p, with sound generated by Kling V3, hushed room tone included. Enough for a story, a pre-roll or an animated product page, for under three dollars for the whole run.
Describe your product
In the Reference node: the name, the material, the colour, what has to stay legible, and two or three photos from different angles. The more precise the card, the less the model improvises.
Set the camera
Focal length, aperture, angle and movement, in the Camera node. The template starts at 85 mm, f/1.4, low angle and slow dolly in: a classic advertising starting point, to change according to the effect you want.
Generate the still
The Image node produces the lit scene with your product in it. This is the moment to regenerate: a failed scene costs a few cents, a failed video costs ten times that.
Run the eight seconds
The approved still becomes the first frame of the video. The movement written in the Text node and the camera settings drive the animation, and the sound is generated along with the picture.
The same canvas, tuned for different needs: each variation is two or three lines of prompt away.
The direct use case: one object, studio lighting, eight seconds. Start from your existing product photos, or make them first with the product photo workflow, and save the price and the tagline for the edit, never for the generation.
Eight seconds in 16:9 lands just under the non-skippable formats. The movement has to start on the first frame: a shot that takes two seconds to move has already lost half its audience.
Switch the Image node and the Video node to 9:16 and the rest of the workflow stays put. Regenerating in native vertical always beats cropping a 16:9, which cuts exactly the area the product sits in.
With no text and no promise, eight seconds looping at the top of a product page often beats a still image. Bring the movement down to a very slow dolly: a quiet loop stays watchable, a nervous one wears out in three passes.
A prototype, packaging still in design, a colour variant to settle: the product card accepts renders instead of photos. Useful for showing a client three directions before committing to an actual shoot.
The Camera node and the product card stay in place, only the scene description changes: dark stone, white marble, outdoors at sunset. Three generated scenes, one camera setup, so a series that actually holds together.
Writing 85 mm into the scene text hands the model one more word to interpret. The Camera node goes to the image and to the video in a form both models read the same way. That is the point of the node, and the camera direction guide has the detail.
The video costs most of the workflow and inherits every flaw in the still. A label that is unreadable in the image will be unreadable and blurred for eight seconds. Regenerate the image until it stands on its own.
A fast dolly, a full rotation or a hard zoom deform the product along the way. Slow moves are what advertising actually uses, and they are also what video models hold best.
A price, a logo or a tagline generated into the image comes out distorted half the time, and changes from one render to the next. Text belongs in the edit, where it stays sharp, editable and identical across the whole series.
It lets you write a cinematography brief instead of scattering it through the prompt. Focal length, aperture, angle and movement go to the image and the video at the same time, so both steps describe the same shot. The camera direction guide covers it in detail.
Yes, two or three are enough, taken from different angles. That is what guarantees the bottle in the video is yours, with its shape, its label and its material, and not a lookalike.
Yes, Kling V3 produces the audio track along with the picture, from the atmosphere described in the movement instruction. Nothing to lay on top if that atmosphere is enough.
Under three dollars for the scene and the 1080p video with sound. The price shows on each node before you click, and the scene alone costs a few cents, which lets you try several settings before animating.
Yes, the renders are yours and they are made for advertising. The caution is on what you feed in: a third-party brand, a real person's face or music under copyright stay subject to their own rules.
The UGC video puts a person on screen talking about your product, with their voice. Here there is nobody: it is a brand advertising shot, directed at the camera level, with the product alone on set.
Yes, by duplicating the Image node and the Video node with different camera settings, then cutting the shots together. Keeping the same product card from one shot to the next is what holds the object consistent.
Eight seconds per video generation. Beyond that you chain shots rather than stretch one: that is what a real ad does too, where a single shot rarely runs past four seconds.
Make a cinematic ad with AI
Use this workflowFree account, no card required. The workflow opens pre-filled in your canvas.
Seedream 5 Pro
Kling V3