From Blender to AI video
A ten-second token animation led me to separate the work into three parts: Blender for motion, Dreamina for exploring metal finishes, and human review for brand accuracy.
The most useful step was not a longer prompt. It was locking the camera, positions and timing in a blockout, then giving each reference a clear job.
One animation, three different problems
The aim was to raise six tokens in turn at a fixed display position, lower each one, then shift the row one place to the left. The camera needed to stay still, with the brand text above the action.
The challenge was maintaining the motion, each token's identity and a consistent silver-metal finish at the same time. Those problems needed different controls; asking for a more metallic look could not solve all three.
Use the blockout to define the motion
With help from Codex and GPT, I wrote Blender Python using bpy to build the coins, base, camera and keyframes. The blockout turned a description into a motion sequence I could inspect, one decision at a time.
It also clarified the mechanism: a conveyor-style display, with a fixed presentation point and a moving row underneath. Locking the camera, occlusion and timing meant the generation stage did not have to invent everything at once.
An identity texture is not an identity lock
In the first generation, the raised token looked metallic while the row beneath still resembled the blockout. I added identity textures to each token in Blender so the marks were present from the first frame.
The reference became clearer, but AI could still redraw logos, reorder tokens or mix frames from different references. A priority written in a prompt is not the same as a locked object property.
A longer prompt does not guarantee more control. Reducing conflicts between references makes the problem easier to locate.
Review visual quality and accuracy separately
Blender V6 kept the order and patterns more stable, but its materials, lighting and geometry did not reach the intended finish in this test. Dreamina came closer to the desired metal look, while brand accuracy still needed a frame-by-frame check.
The division of work became clearer: Blender for cameras and motion, Dreamina for visual exploration, and a controlled post-production step for exact logos, typography and final brand review.
A five-step Blender and AI video workflow
- 01
Define the target. Start with a clear image of the finish you want.
- 02
Lock the motion. Fix the camera, positions, occlusion and timing in Blender.
- 03
Establish identity. Make each object recognisable from the first frame.
- 04
Separate the references. Give motion, materials, logos and environment their own roles.
- 05
Review in separate passes. Assess the finish, then check logos, order and continuity.
