4K Upscale settings for AI video content
Mastering post-generation upscaling parameters and character-integration prompting enables AI filmmakers to achieve clean, cinematic 4K deliverables from low-resolution generation drafts.
What this lesson covers
Joey discusses the evolving landscape of AI filmmaking and demonstrates practical workflows for upscaling AI video footage to 4K resolution using Topaz. He also outlines how to seamlessly integrate generated AI characters into real-life studio backgrounds by leveraging Claude for frame-locked prompts.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 02:43 ↗
Utilize the Iris enhancement model in Topaz for optimal facial detail when upscaling AI video footage, using Proteus as an alternative.
- 03:32 ↗
Target 3840x2160 resolution for 16x9 footage while adding minor digital noise (values 2-3) to avoid unnatural AI sharpness.
- 05:24 ↗
Configure custom upscale dimensions of 5040x2160 with a constant bitrate set to 24 when upscaling 21x9 ultra-widescreen AI video.
- 06:13 ↗
Combine live-action background plates and character reference sheets in Claude with explicit frame-locking prompts to composite characters into real scenes.
Workflow outlined in the video
- Generate draft video renders at 720p resolution to conserve platform generation credits prior to upscaling.
- Import 16x9 footage into Topaz Video AI and select the Iris model with an output resolution of 3840x2160.
- Inject 2 to 3 units of digital noise and optional low-density grain (size 1.5–1.8) to give upscaled clips cinematic texture.
- Switch codec bitrate settings from dynamic to constant at 24 when upscaling 21x9 aspect ratio shots to 5040x2160.
- Pass background plates and character reference sheets into Claude to construct precise spatial and frame-locked prompts for scene compositing.
Before you use this workflow
These notes describe the source video at its publication date. Model access, pricing, connectors and interfaces may have changed. Check the original source and the provider’s current documentation before installing an add-on, connecting an account or spending credits.
ReelStack has not independently tested this workflow. Preview one representative shot and check motion, continuity and output quality before applying it to a full production. No result, cost saving or model capability is guaranteed.
Explore the tools
How to interpret the numbers
21,481 views captured 24 October 2026. ReelStack recommendation score: 38/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 65% performance: 0.80× current views vs the channel's recent median; same-age history is not yet available. View-evidence factor 96% (views / (views + 1,000))
- 25% freshness: 0/100 with a 14-day half-life
- Momentum pending: collecting daily snapshots; its 20% weight goes to observed performance, not free points
- 10% engagement: 100/100; likes + 4× comments, smoothed with a 500-view neutral prior