NevermindNilas/TheAnimeScripter

A high-performance tool for video upscaling, interpolation, depth estimation, and more. Available as a CLI and Adobe Extension.

🎬 The Anime Scripter (TAS) – What It Is

The Anime Scripter (TAS) is a Windows‑focused toolkit that lets creators apply a range of AI‑powered video‑enhancement operations to anime or any other footage. It bundles several state‑of‑the‑art models (upscalers, frame‑interpolators, denoisers, depth generators, object detectors, etc.) and offers three ways to use them:

Edition	How you run it	Who it’s for
TAS‑Standalone	A single‑click installer with a graphical UI; the backend is shipped inside the installer, so you don’t need Python or any command line tools.	Users who want a simple “open‑app‑drag‑video‑run” experience.
TAS‑AdobeEdition	An After Effects plugin that runs the same backend from inside AE.	Video editors who stay inside Adobe’s workflow.
CLI	`main.py` / `tas` command line with every parameter exposed.	Power users, script writers, and anyone who wants to chain operations in batch.

🚀 Core Capabilities (as described in the README)

Category	Features
Video Enhancement	• Motion interpolation (smooth frame‑rate up‑conversion) • 2× AI upscaling (multiple models) • Motion‑blur synthesis via weighted frame blending • Smart deduplication to drop redundant frames
Advanced Editing	• Automatic background‑foreground segmentation (rotobrush‑like) • Depth‑map generation for 3‑D effects • YOLOv9 object detection with masks and class labels • Video stabilization using SuperPoint feature matching
Workflow Optimisation	• Desktop GUI with presets, live progress, theming • After Effects integration • Model chaining – run several effects in a single pass • In‑memory frame handling to avoid extra disk I/O
AI Flexibility	• Supports CUDA, TensorRT, DirectML, OpenVINO back‑ends • You can drop in your own ONNX/Spandrel models • Restoration options: denoise, de‑jpeg, sharpen, line‑enhancement

📦 What’s Inside – Models

The README lists the models that TAS can download on demand. They are grouped by task and each shows which hardware back‑ends are supported.

Upscaling – ShuffleCugan, Fallin (Soft/Strong), Span, OpenProteus, AniScale 2, RTMOSR, Saryn, Gauss, custom Spandrel, NVIDIA Maxine VSR.
Interpolation (RIFE) – Versions 4.6 up to 4.25 (including lite and heavy builds) and a community‑modified Rife_Elexor.
Restoration – SCUNet, NAFNet, DPIR, Real‑Plksr de‑jpeg/de‑h264, Anime1080fixer, FastLineDarken, GaterV3, deepDeband‑f, NVIDIA Maxine denoise/deblur.
Depth Maps – Depth‑Anything‑v2 (small) and the original implementation.
Object Detection – YOLOv9 Small/Medium/Large.
Stabilization – SuperPoint feature‑matching (CUDA only).

🛠️ Getting Started (from the README)

1. TAS‑Standalone (recommended for most users)

Download the TAS‑<version>‑Standalone‑Windows‑Setup.exe from the Releases page.
Run the installer – it bundles the AI backend, so no Python or extra libraries are needed.
Launch the app, load a video, pick the desired effects, and hit Run.
- The first time you use a particular model, TAS will download it, which adds a one‑time delay.

2. After Effects Plugin

Grab the TAS‑AdobeEdition zip from the releases.
Extract it somewhere convenient.
Follow the linked installation tutorial (zxp‑installation) to add the plugin to After Effects 2022+.
Use the new TAS panel inside AE to apply the same AI effects without leaving the editor.

3. Command‑Line Interface

Installation (Windows) – one‑liner PowerShell:
```
iwr -useb https://tas.nevermindnilas.dev/install.ps1 | iex
```
This creates a TheAnimeScripter folder and optionally adds it to your PATH.

Installation (Linux/macOS) – clone the repo, then install Python requirements:

python -m pip install -r requirements.txt -r extra-requirements-windows-lite.txt   # or the macOS profile

Run tas --help (or python main.py -h) to see all options. The README points to a PARAMETERS.MD file for the full list.

🎯 Typical Use Cases

Scenario	How TAS Helps
Anime up‑scaling	Choose an upscaler model (e.g., ShuffleCugan) and run a 2× upscale in the GUI or via CLI.
Smooth slow‑motion	Use the RIFE interpolation models to raise the frame‑rate (e.g., 30 fps → 60 fps) before exporting.
Clean up compression artifacts	Apply a restoration model such as SCUNet or DeJpeg to reduce noise and blockiness.
Add depth for 3‑D post‑processing	Generate a depth map with Depth‑Anything‑v2, then feed it into compositing tools for parallax effects.
Automated batch processing	Write a shell script that calls `tas` with a chain of models (e.g., dedup → upscale → denoise) on a folder of videos.
Integrate into motion‑graphics pipelines	Use the After Effects plugin to run AI upscaling directly on a composition layer, keeping the workflow inside AE.

📋 System Requirements (from the README)

OS – Windows 10/11 (64‑bit). macOS is supported for the CLI only (Apple Silicon via the MPS profile). No native Linux GUI is mentioned.
GPU – A modern NVIDIA RTX card gives full CUDA/TensorRT support. Older NVIDIA GPUs, Intel integrated graphics, or AMD GPUs fall back to DirectML or OpenVINO where indicated.
Dependencies – The GUI installer bundles everything. The CLI needs Python 3, ffmpeg (installed via Homebrew on macOS), and the appropriate extra‑requirements‑*.txt for your hardware.

🙏 Community & Contributions

The project lists a fairly extensive set of contributors, most of whom provide the underlying AI models (ShuffleCugan, RIFE, YOLOv9, Depth‑Anything, etc.) or the video‑processing libraries (FFmpeg, yt‑dlp, PySceneDetect). The README also includes links to a Discord server, a star‑history chart, and a contributors‑rock image.

✅ Bottom Line

The Anime Scripter is a genuine, feature‑rich AI video‑enhancement suite aimed at creators who work with anime or other animated content. It bundles a variety of open‑source AI models, offers a no‑code desktop UI, an After Effects plugin, and a full‑featured CLI for automation. All of this is clearly described in the repository’s README, making it an on‑topic example of an AI‑driven media‑processing tool.