Rendering
What is Rendering?
Rendering is the process of turning data into a finished visible image: whether a 3D programme calculating how light hits surfaces, or an AI model converting its internal calculations into the image you see.
At a glance
- Also known as
- Image rendering3D rendering (in CGI contexts)Inference (in AI contexts)Generation (colloquially)
- Used for
- Converting 3D scene data into final photorealistic or stylised imagesProducing final pixel output from AI model computationsThe terminal production step in CGI, animation, VFX, and AI generation
- Common tools
- Arnold, v-ray, RenderMan (3D offline rendering)Unreal engine, unity (real-time rendering)Stable diffusion, midjourney, runway (AI rendering)Blender cycles / EEVEE
- Related terms
- Diffusion modelSamplingPost-processingCompositingRay tracingLatent space
Ready to create?
Direct scenes, design characters, and ship full films
All-in-one AI creative platform with simple, transparent pricing, no speed throttles, and an infinite Canvas for max creativity.
How it compares
Compared with related concepts
Rendering and compositing are sequential stages of the visual production pipeline that are often confused. Rendering produces the initial image or sequence of frames from scene data or model inference. Compositing takes rendered elements and combines, layers, and integrates them: adding VFX elements to live footage, combining foreground and background renders, applying colour grades. Rendering produces the raw visual components; compositing assembles them into the final image. In AI generation, post-processing serves a similar role to compositing in traditional pipelines.
Think of it like…
Rendering is like developing a photograph in a darkroom: the shoot (or scene setup, or AI generation) captured all the information needed, but rendering is the chemical or computational process that actually produces the visible image from that information. Until rendering is complete, there is no image: only the data from which an image can be produced.
Pro tip
When evaluating AI generation models for production use, render quality ( not just subject accuracy ) should be a key assessment criterion. Specifically look at how the model renders fine detail (hair, fabric weave, surface texture), how it handles specular highlights and shadow transitions, and whether it maintains consistent rendering quality across the frame rather than concentrating detail in the focal subject and degrading toward the edges. These rendering quality differences between models are often as significant as their subject and style differences.
Types and variations
- Ray tracing renders images by simulating the path of individual light rays through a scene, producing highly accurate shadows, reflections, and global illumination at significant computational cost.
- Rasterisation projects 3D geometry directly onto the 2D image plane, enabling real-time rendering at the cost of physical accuracy.
- Path tracing is a probabilistic extension of ray tracing used in offline film production for maximum physical accuracy.
- AI diffusion rendering generates images through an iterative denoising process guided by learned model parameters.
- Real-time rendering prioritises speed for interactive use; offline rendering prioritises quality for final deliverable production.
- Hybrid rendering combines rasterisation with ray-traced lighting for modern games and interactive applications.
Ready to make your first scene in Morphic?
Try MorphicCommon use cases
- Rendering is the final production step in all computer-generated imagery: animated film, visual effects, architectural and product visualisation, video game graphics, real-time interactive environments, and AI image and video generation.
- Understanding rendering quality requirements: the resolution, frame rate, and visual fidelity expected of the final deliverable: informs decisions made at every earlier stage of the production pipeline, from scene complexity to model selection to post-processing requirements.
Ready to create?
Direct scenes, design characters, and ship full films
All-in-one AI creative platform with simple, transparent pricing, no speed throttles, and an infinite Canvas for max creativity.
FAQs
Rendering is the computational process of converting underlying data — 3D scene geometry and lighting, or an AI model's internal representations: into the final pixel values of a viewable image or video frame. It is the terminal step in any visual generation pipeline, producing the actual image from the information that precedes it.
Real-time rendering must produce frames fast enough for interactive use: typically 30 to 120 frames per second on consumer hardware. It uses techniques like rasterisation that prioritise speed over physical accuracy. Offline rendering has no speed constraint and can use computationally expensive algorithms like path tracing that produce physically accurate results but may take minutes, hours, or days per frame. Film VFX uses offline rendering; video games use real-time rendering.
Ray tracing is a rendering algorithm that simulates the physical behaviour of light by tracing the path of individual light rays from the camera through the scene, calculating how they interact with surfaces, materials, and light sources. It produces highly accurate reflections, refractions, and shadows but is computationally expensive. Modern ray tracing hardware acceleration has made real-time ray tracing practical in consumer GPU hardware, though full path-traced rendering remains primarily an offline technique.
In AI generation, rendering refers to the inference process: the series of computational steps through which the model transforms its internal representations (latent vectors, noise fields, weight-conditioned activations) into the final visible image. AI rendering quality is determined by model architecture, training data, and inference parameters. The term is used both technically (the inference computation) and colloquially (the act of generating an image).
Rendering artefacts are visual errors or anomalies produced by the rendering process: unintended visual defects that reveal the limitations of the rendering algorithm or system. In 3D rendering, artefacts include shadow acne, fireflies, polygon clipping, and aliasing. In AI generation, artefacts include blurring, distortion, anatomical inconsistencies, tiling patterns, and implausible physics. Both types result from the rendering process producing outputs that do not accurately represent the intended scene or generation.
Render time directly affects production cost, iteration speed, and creative flexibility. Long render times reduce the number of iterations a team can complete in a given time budget, limiting the ability to explore variations and correct problems. In AI generation, generation time (the AI equivalent of render time) affects workflow efficiency. Faster models allow more rapid iteration; slower models with higher quality may be appropriate for final deliverable production but impractical for exploration and development.
A render farm is a network of connected computers or servers dedicated to rendering, distributing the computational work of producing images or video frames across many machines simultaneously. Render farms enable productions to complete renders that would take months on a single machine by distributing the work across hundreds or thousands of processors running in parallel. Cloud-based render farms allow productions to access large-scale rendering capacity on demand without maintaining physical infrastructure.
In diffusion model AI generation, step count refers to the number of denoising iterations performed during the inference process. More steps allow finer, more incremental refinement of the image as noise is progressively removed, generally producing higher detail and coherence. Fewer steps produce results faster but with coarser resolution of fine detail and potentially more artefacts. Most models have a range within which step count meaningfully improves quality, beyond which additional steps produce diminishing returns.
