AI-generated visual content is moving fast. In professional settings, though — pharmaceutical, technical training, scientific communication — speed on its own is not enough.
The challenge here is a different one: control, repeatability and visual coherence.
At Inneva Pharma we have spent some time integrating new AI technology into 3D pipelines that already work, combining tools such as Gaussian Splatting, ComfyUI and classic 3D pipelines to build scenes and video with a high degree of precision.
This is not about replacing existing processes. It is about improving them.
Start with structure: the 3D animatic as a base
Our starting point is always a 3D animatic, usually built in Blender.
Even though the models are generic and carry no final detail, this phase is critical, because it settles:
- framing
- camera movement
- narrative pace
- the spatial structure of the scene
The visual narrative is therefore closed from the outset, which matters in training, scientific or regulatory projects, where improvising visually is not an option.
Styleframes: fixing the visual language
Once the animatic is defined, we produce one or more styleframes that establish the visual language of the project:
- materials
- lighting
- atmosphere
- graphic style
These styleframes act as a solid visual reference and head off one of the most common problems with AI-generated content: the lack of coherence between shots or between iterations.
ComfyUI and Gaussian Splatting as the production layer
With the animatic and the styleframes settled, we bring ComfyUI into the pipeline to generate the final scene as video.
What makes this approach work is that:
- the camera movement is respected exactly
- the composition does not distort
- the spatial structure stays stable
That allows advanced visual results without giving up the control that traditional 3D provides.
On top of that, working with layers and per-element identifiers (ID passes) lets us act on specific parts of the scene:
- change textures
- adjust materials
- modify specific elements
without redoing the whole shot.
What it gives you in professional work
This hybrid approach brings clear advantages to real projects:
- Faster iterations, without breaking the scene
- Greater visual control, which is decisive in regulated environments
- Reuse of shots and cameras, even when the style changes
- Lower cost and shorter timelines, without sacrificing quality
For pharmaceutical work, medical training or scientific communication, this is a significant step up from purely generative flows or from traditional renders that are too rigid to iterate on.
AI and 3D: not a choice, a combination
The question is no longer whether to use AI or 3D.
The right question is where each one adds value.
At Inneva Pharma we use:
- 3D to structure, define and control
- AI to accelerate, iterate and enrich visually
When the two are integrated coherently, the result is content that is more solid, more efficient and better matched to what our clients actually need.
