AI Enhance

Tab: AI Enhance — Upscale & denoise · Plugin: Cinematic Lab

Three separate things, in increasing distance from Unreal: viewport upscaling, offline path-tracer denoising, and a hand-off to third-party binaries you install yourself.

Cinematic Lab ships none of these tools. This tab configures Unreal’s own upscalers and writes a recipe plus launcher for external ones you provide.

The AI Enhance tab: real-time upscaler, path-tracer denoiser and the post-export pipeline.

Real-time upscaler

Pick an Upscaler and a Quality level, then:

• ✓ Apply to Viewport. Sets ScreenPercentage plus the chosen vendor’s console variables.
• Disable All. Resets ScreenPercentage to 100 and disables DLSS, FSR and XeSS.

Setting the upscaler and denoiser options in the AI Enhance tab.

Availability is detected, not assumed

The panel reports what your machine and project can actually do:

Available — TSR: ✓  ·  DLSS: <state>  ·  FSR: <state>  ·  XeSS: <state>

TSR is Unreal’s own temporal super resolution and is always available. DLSS, FSR and XeSS depend on vendor plugins being installed and on hardware support, so their state is probed rather than promised.

Offline (Path Tracer)

For path-traced renders, a PT Denoiser choice:

Option	What it is
DLSS Ray Reconstruction	Replaces the spatial denoiser
DLSS Super Resolution	Offline upscale — render half-resolution, output at target

These affect the path-traced render path, not the raster viewport.

Post-export AI pipeline

The part that leaves Unreal. Three external tools, each with a path field:

Tool	Notes
Topaz Video AI	Path to the Topaz CLI binary — license required
Real-ESRGAN	Path to realesrgan-ncnn-vulkan.exe — free
RIFE (frame interp)	Path to rife-ncnn-vulkan.exe — free

The panel labels the licensing plainly on each field. Real-ESRGAN is described as the “free fallback to Topaz” rather than pitched as equivalent.

Writing a recipe

✏️ Write AI Pipeline Recipe writes <OutputDir>/eqlabs_ai_pipeline.json describing the Topaz, ESRGAN and RIFE settings. Status: Last recipe: <path>, or “No recipe written yet.”

The recipe is a plain JSON file. It documents the intended pipeline whether or not you ever run it from here — useful when the actual processing happens on a different machine or a farm.

An AI pipeline recipe written to the output directory.

Running a step

Button	What it runs
▶ Topaz	Topaz CLI on the configured frame pattern
▶ Real-ESRGAN	Real-ESRGAN x4 on the configured frame pattern
▶ RIFE	The RIFE frame interpolator on the configured frame pattern

While a step runs the status reads ● <tool> running…; when it finishes, <tool> <result> · exit <code> · <seconds> s. The exit code is reported rather than a green tick — if the external tool failed, you see its actual exit code.

With nothing run yet: “No handoff yet.”

Notes

• Nothing here is bundled. Topaz needs a license you buy; Real-ESRGAN and RIFE are free downloads. Cinematic Lab stores paths, writes a recipe, and shells out.
• The upscaler section touches project state. Apply to Viewport sets ScreenPercentage and vendor CVars; Disable All puts them back. If the viewport looks soft after experimenting, that is the button.
• Availability is probed per machine. A ✗ next to DLSS means this machine or project cannot use it — installing the vendor plugin is the fix, and the tool does not pretend otherwise.
• Exit codes are surfaced. A failed external run reports its code instead of being swallowed.