EmbyVision OCR - CPU Based AI Credits Detector for Emby (Docker)

[EncryptedCity 20]

EmbyVision OCR is a standalone OCR service designed to improve credits detection for Emby libraries. It runs separately from Emby in Docker and analyzes video frames using OCR to help identify end credits.

Note: This is not an Emby plugin. It is a standalone companion service designed to work with Emby Credits Detector integrations.

Features

• Lightweight design with minimal resource requirements, approx. 1.89GB pulled

• Consuming approx. 345.2MiB - 3.5GB RAM while in use.

• CPU-only OCR processing

• No GPU required

• No NVIDIA drivers required

• No CUDA setup required

• AI-powered OCR using PaddleOCR

• Designed for Emby Credits Detector workflows

• Docker-based deployment

• No database required

• No persistent appdata storage required

• Self-contained container image

• Simple Docker Compose installation

• Optimized for self-hosted media servers

• Optimized for fast CPU-only OCR processing

 

Container Image

The official CPU-only container image is available through GitHub Container Registry.

This release is CPU-only and does not include GPU acceleration.

GitHub Repository

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

Docker Installation

Docker Run

docker run -d \
  --name embyvision-ocr \
  --restart unless-stopped \
  -p 8000:8000 \
  ghcr.io/encryptedcity/embyvision-ocr:latest

Docker Compose Installation

Create:

docker-compose.yml

 

Add:

services:
  embyvision-ocr:
    image: ghcr.io/encryptedcity/embyvision-ocr:latest
    container_name: embyvision-ocr
    restart: unless-stopped
    ports:
      - "8000:8000"

 

Start the container:

docker compose up -d

 

 

This image does not require:

• NVIDIA Container Toolkit
• CUDA libraries
• GPU passthrough
• GPU device mappings

Compatibility

• Tested with Docker deployments

• Runs well on Unraid

• CPU-only operation makes it compatible with systems without dedicated AI hardware

• Suitable for self-hosted Emby servers

 

Edited July 31 by EncryptedCity
Banner

[enterpr1se 3]

Hi,

Is the Docker image ghcr.io/encryptedcity/embyvision-ocr:latest still available?
I tried pulling it, but it returned a "404 Not Found" error.

[EncryptedCity 20]

Yes, it is still available I just pulled it. ghcr.io/encryptedcity/embyvision-ocr:latest

[EncryptedCity 20]

Hi, information for everyone. the new plugin can be downloaded here. https://github.com/encryptedcity/EmbyVision-Credits 

The message below is no longer current. Please use the correct plugin. EmbyVision Credits plugin.

Note: This is not an Emby plugin. It is a standalone companion service designed to work with Emby Credits Detector integrations.

[enterpr1se 3]

2 hours ago, EncryptedCity said:

Yes, it is still available I just pulled it. ghcr.io/encryptedcity/embyvision-ocr:latest

Hi, I just tried pulling it again, but I'm still getting the error. Could you double-check if the repository visibility or image permissions are set to public?

[EncryptedCity 20]

41 minutes ago, enterpr1se said:

Hi, I just tried pulling it again, but I'm still getting the error. Could you double-check if the repository visibility or image permissions are set to public?

Retry pulling now. ghcr.io/encryptedcity/embyvision-ocr:latest it should work. They have been changed to public.

Edited September 21 by EncryptedCity

[enterpr1se 3]

17 hours ago, EncryptedCity said:

Retry pulling now. ghcr.io/encryptedcity/embyvision-ocr:latest it should work. They have been changed to public.

Thanks! I can pull the Docker image now. However, I’m having trouble starting the container—my NAS only has 12GB of RAM, which might not be enough.

[EncryptedCity 20]

The container needs time to download and initialize. Check the logs, I would not suspect the 12GB ram is your problem.  docker logs -f embyvision-ocr

Edited September 22 by EncryptedCity

[EncryptedCity 20]

This is what the logs should look like embyvision-ocr  | INFO:     Uvicorn running on http://0.0.0.0:8000 It also shows the ram usage. Approx 350-500 MiB during startup. Here is a video of pulling the container and what to expect.