Topaz Labs brought its latest upscaling and restoration tools to NAB 2026, with Principal Creative Technologist Ross Shain walking through how the platform handles the cleanup, enhancement, and upscaling work that consumes hours of post-production time.
Key takeaways:
AI-powered upscaling handles resolution increases from SD to 4K and beyond with detail preservation that avoids the soft look of traditional scaling
Restoration models address film grain, scratches, noise, and compression artifacts in a single pass
Batch processing lets teams run hundreds of clips through the same pipeline without manual intervention
Upscaling That Preserves Detail
Traditional upscaling methods like bicubic interpolation stretch pixels and fill in the gaps with averaged values. The result looks soft. Topaz AI uses neural networks trained on pairs of low-resolution and high-resolution images to predict what the missing detail should actually look like.
Shain demonstrated footage upscaled from 1080p to 4K where fine textures like fabric weave, skin detail, and hair strands remained sharp rather than smoothed out. The model adds plausible detail based on its training, which means the output looks like it was originally shot at the higher resolution rather than stretched after the fact.
The practical application: archival footage, stock footage purchased at lower resolutions, and older camera files that need to match modern delivery specs. Instead of reshooting or accepting visible quality loss, producers run the footage through Topaz and deliver at the target resolution.
Restoration for Damaged Footage
The restoration side handles footage that has degraded over time or was captured under poor conditions. Shain showed examples of film transfers with visible scratches, grain, and color fading that were cleaned up in a single processing pass.
The restoration models target specific damage types. Film grain reduction distinguishes between intentional cinematic grain and unwanted noise from underexposure or poor transfer conditions. Scratch removal fills in damaged frames using context from surrounding frames. Compression artifact cleanup addresses the blocky artifacts that appear in heavily compressed video files.
Each model can be applied independently or combined. A restoration pipeline for archival film might run grain reduction, scratch removal, and color correction sequentially, with the user adjusting the intensity of each step.
Batch Processing for Production Volumes
The batch processing system lets teams define a pipeline of enhancement steps and apply it to hundreds of clips automatically. Instead of processing each file individually, users set their parameters once and let the software work through the queue.
This matters for productions dealing with large volumes of footage: documentary teams working with archive material from multiple sources, post houses delivering restored versions of film libraries, and content teams standardizing mixed-resolution footage into a consistent output format.
Shain noted that the batch system preserves file metadata and supports common professional formats including ProRes, DNxHR, and EXR sequences.
What Is Coming
Shain hinted at upcoming features focused on AI-driven color grading assistance and real-time processing capabilities, though specific timelines were not shared. The direction is toward integrating Topaz tools directly into editorial workflows rather than requiring a separate processing step.
Topaz AI handles upscaling, restoration, and batch enhancement of video footage so post teams can deliver clean high-resolution output without manual frame-by-frame work.
