St6.Lapidary 🍂

🍂 Showcase

Drag the handle to compare.

Real frames pulled straight out of an AE timeline. Slide between source and St6 output — no re-encode, no compression tricks.

🪞 Upscale Real-CUGAN-Pro (NCNN-Vulkan) · 2× · Lapidary's color boost on

BEFORE AFTER

⇆

⏱ Interpolation RIFE 4.25-heavy on TensorRT · 16× — re-timed to source duration

BEFORE AFTER

⇆

🍂 What it does

Six tools, one panel.

⏱

Frame interpolation

RIFE 4.25-heavy on TensorRT runs at ~200 fps. Slow-motion, ensemble, scene-cut detection — any factor from 1.5× to 40×.

⬆

AI upscaling

Real-CUGAN-Pro, Real-ESRGAN, Waifu2x, ShuffleCugan, SPAN. 2×/3×/4× with TTA, denoise, and an optional color boost.

🌊

Depth maps

Depth-Anything-V2 with 22 OpenCV colormaps + RGB-guided edge sharpening. Native resolution for 2.5D parallax + shaders.

⌇

Smart queue

Drag-reorder, pin-to-top, cancel-all. Color-coded status with live phase + percent + ETA. Survives AE crashes.

🎨

Themes & a11y

Autumn, Midnight, Sakura, Forest. Custom-accent picker, reduced-motion mode, color-blind palette, output filename templates.

🔔

Notifications

JRPG-style sound bank when renders finish — Coin, Level Up, Quest, Item Get. Discord rich presence with the 🍂 logo.

🍂 Community

Edits in the wild.

A rolling feed of edits made with St6.Lapidary. Want yours featured? Drop it in the Discord #showcase channel and I'll add it here.

+

Submit your edit

drop it in #showcase on Discord

🍂 Live stats

From real installs.

Anonymous, opt-in usage telemetry. Refreshed every few minutes from the Cloudflare backend.

Renders · 7d

—

Success rate · 7d

—

Active installs · 7d

—

Failed renders · 7d

—

Daily volume · last 30 days

GPU breakdown

Top upscale models · 30d

Top interpolation models · 30d

—

🍂 Get St6.Lapidary

Install once, render forever.

A signed ZXP that bundles every model and dependency. Drop it on ZXPInstaller and you're rendering in two minutes.

🪟

Windows · Latest release

Signed ZXP · CUDA / TensorRT / DirectML / NCNN

→

🍂 FAQ

Asked & answered.

Tap any question to reveal St6's reply.

Curious user Does it work with my GPU?

St6 🍂 RTX 20/30/40/50 → CUDA + TensorRT (fastest). GTX 16 → CUDA. GTX 10 (Pascal) and AMD → DirectML or NCNN. Intel iGPUs/dGPUs → NCNN-Vulkan. Each model in the panel shows which backends it supports.

Curious user Is it really free?

St6 🍂 Yes — no paywall, no watermarks, no time limits. Telemetry is opt-in and asks before sending anything. Use it in your own work however you want; just don't redistribute or repackage it.

Curious user Do I need Python or any extra installs?

St6 🍂 No. The ZXP bundles everything — PyTorch, TensorRT, ffmpeg, Real-CUGAN binaries, all of it. On first run St6 unpacks the engine into %LOCALAPPDATA%\\st6.lapidary and never touches your system Python.

Curious user Can I use my own models?

St6 🍂 Yes — drop any compatible .pkl / .pth / .onnx / .engine into the matching backend/models/<name>/ folder and it shows up in the panel's model dropdown.

Curious user Can I turn off the color boost?

St6 🍂 Yes — Advanced Settings → Pipeline → toggle Lapidary color boost. By default the final render gets a +10% saturation + 0.93 gamma pass that pops anime colors. Flip it off if you want raw model output.

Curious user What does telemetry collect?

St6 🍂 Model + factor, render duration, source-resolution bucket (HD / FHD / 4K — never the actual dimensions), GPU vendor, app version, OS family, and an error class on failure. Source clip names, file paths, and your IP never leave your machine.

Curious user Will there be new features?

St6 🍂 Yes — new models, backends, and quality-of-life features land regularly. Follow GitHub releases or join the Discord for dev builds and early peeks.

AI video, polished.