🚀 EmbyVision OCR (GPU Edition) — NVIDIA GPU Accelerated OCR for Emby Credits Detector

[EncryptedCity 17]

Here's a polished Emby Community forum post that would fit well in the Tools & Utilities section. It explains what the project is, who it's for, and gives users enough information to get started without overwhelming them.

---

EmbyVision OCR (GPU Edition) — NVIDIA GPU Accelerated OCR for Emby Credits Detector

Hello everyone!

I'm excited to share EmbyVision OCR (GPU Edition), a standalone OCR microservice designed specifically to work alongside the Emby Credits Detector plugin.

EmbyVision OCR uses PaddleOCR with NVIDIA CUDA acceleration to provide fast, AI-powered text recognition from video frames, helping improve end credit detection while significantly reducing OCR processing time compared to CPU-based solutions.

---

What is EmbyVision OCR?

EmbyVision OCR is a lightweight REST API that performs Optical Character Recognition (OCR) on video frames.

Instead of relying on traditional OCR engines, it uses PaddleOCR's deep-learning models running on your NVIDIA GPU for improved speed and recognition accuracy.

The service runs independently in Docker and integrates directly with the Emby Credits Detector plugin.

---

Features

NVIDIA GPU accelerated OCR

CUDA + cuDNN support

AI-powered PaddleOCR engine

Fast OCR processing

Designed specifically for Emby Credits Detector

Docker deployment

Persistent AI model storage

Automatic model downloads

REST API

Self-hosted

Privacy focused

Works great on Unraid

---

Requirements

This edition requires:

• NVIDIA GPU

• NVIDIA Driver

• NVIDIA Container Toolkit

• Docker

• Emby Server

• Emby Credits Detector plugin

• Docker Image Sizes

  CPU Edition:
  • Image: ghcr.io/encryptedcity/embyvision-ocr:latest
  • Size: 1.89 GB
  • Requirements: Any Docker host
  • GPU: Not required
  • CUDA/NVIDIA Runtime: Not required

  GPU Edition:
  • Image: ghcr.io/encryptedcity/embyvision-ocr-gpu:latest
  • Size: 8.34 GB
  • Requirements: NVIDIA GPU + Container Toolkit
  • CUDA acceleration enabled

---

Not Supported

This edition will NOT work with:

• AMD GPUs

• Intel GPUs / Arc

• CPU-only systems

If you do not have an NVIDIA GPU, please use the CPU Edition instead.

CPU Edition:

https://github.com/encryptedcity/embyvision-ocr

---

Installation

The easiest method is Docker Compose.

services:

  embyvision-ocr-gpu:

    image: ghcr.io/encryptedcity/embyvision-ocr-gpu:latest

    container_name: embyvision-ocr-gpu

    restart: unless-stopped

    runtime: nvidia

    ports:
      - "8001:8001"

    volumes:
      - /mnt/user/appdata/embyvision-ocr-gpu/models:/root/.paddleocr

    environment:
      - NVIDIA_VISIBLE_DEVICES=all
      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
      - CUDA_VISIBLE_DEVICES=0
      - TZ=America/Edmonton

Start the container:

docker compose up -d

---

Unraid Users

EmbyVision OCR works very well on Unraid.

Requirements:

• NVIDIA Driver Plugin

• Docker enabled

• NVIDIA compatible GPU

Container settings:

Repository

ghcr.io/encryptedcity/embyvision-ocr-gpu:latest

Port

8001

Persistent Storage

Host:
/mnt/user/appdata/embyvision-ocr-gpu/models

Container:
/root/.paddleocr

Runtime

nvidia

Extra Parameters

--runtime=nvidia

Environment Variables

NVIDIA_VISIBLE_DEVICES=all
CUDA_VISIBLE_DEVICES=0
NVIDIA_DRIVER_CAPABILITIES=compute,utility

---

First Startup

The first startup downloads the PaddleOCR AI models automatically.

This only happens once.

Future container starts reuse the downloaded models.

---

Connecting Emby

Inside the Emby Credits Detector plugin, configure the OCR server:

http://YOUR_SERVER_IP:8001

Example:

http://192.168.1.100:8001

Then click Test Connection.

If everything is configured correctly, the plugin will begin using EmbyVision OCR automatically.

---

Privacy

Everything runs locally.

• No cloud services

• No telemetry

• No external APIs

• No accounts

• Your media never leaves your server

---

Source Code

GitHub

GPU Edition

https://github.com/encryptedcity/embyvision-ocr-gpu

CPU Edition

https://github.com/encryptedcity/embyvision-ocr

---

Feedback

This project is still actively being developed.

Feature requests, bug reports, testing feedback, and suggestions are all welcome.

If you give it a try, I'd love to hear how it performs on your hardware and whether it improves credit detection in your libraries.

Happy scanning!

Edited 14 hours ago by EncryptedCity
banner

[ebr 16557]

You may want to remove the AI created pre-amble...