EmbyVision OCR GPU Edition, CPU Edition Credits Detector new plugin: Beta

[EncryptedCity 17]

Hello Everyone, I am happy to announce that there is now a Beta version plugin that will detect credits in TV Shows. I originally posted in tools and utilities however since then I have made several improvements to the plugin. The plugins I posted in the tools and utilities were very early versions of EmbyVision OCR. They do support H264 and HEVC files but the latest version of this plugin now supports AV1 as well with CPU and GPU. I developed the software mostly for myself because of my large av1 library and limited tools available that will detect AV1 files for credits. I decided to share it because it is such a cool development.

I realize that this may not be the only credits detector plugin and you are probably wondering, why you should install my version and test it vs the other? Because watching end credits is boring for most!

The original docker containers for the tool are on GitHub. Pease see the posts in tools and utilities for information on installation and also there is installation instructions inside the plugin for the CPU and GPU docker containers. They use the same plugin.

I have also used them with the original in store plugin for OCR detection. They are on GitHub and they do work with the plugin in the Emby forum however it is recommended to use them with my new plugin as it has been designed to work efficiently with them. This image is the OCR Detection working on AV1 files which is mostly what my library is composed of due to my transcoding them to AV1 with my Tdarr and handbrake method.

It's easy to get started. I quite busy and I am not sure how much support I can offer for this plugin but you can find help in the debug logs and they will help you configure the plugin. it  runs on Default settings for most servers. You will need docker for this. 

Feel free to leave comments and post your experiences with the plugin or ask for help,  if I can't get to answering in time surely other people will be able to help you if you have questions. There is lots of documentation. Please read through it. The software also works well on Unraid.

Thank you. Please watch the video. I uploaded one.

EmbyVisionCredits.dll

Edited 1 hour ago by EncryptedCity
video

[yocker 1764]

That looks like a reskinned EmbyCredits?
Not that i mind, would just like to know for clarification sake.

Edited 1 hour ago by yocker

[EncryptedCity 17]

It’s a fair question. It may look similar because it started from the same general idea and builds on the EmbyCredits concept, but it has been heavily modified and expanded. There are new components, including the OCR service, GPU/CPU OCR support, different detection options, new configuration, and a lot of changes under the hood. The detection engines have also been designed and developed from scratch rather than simply being copied over or rebranded.

I’m a little surprised you were able to reach that conclusion only 23 minutes after the post went up though, since the documentation and changes are fairly extensive. I’d be interested to know which parts you looked at that made it appear to be just a reskin.

[EncryptedCity 17]

EmbyVision OCR GPU Edition, CPU Edition can also be installed alongside the original EmbyCredits plugin, so users can run both independently if they want to compare functionality or migrate at their own pace. You can also install both the CPU and GPU containers simultaneously. 

[EncryptedCity 17]

I wanted to share a quick demonstration of the progress so far. The credits detection is working extremely well and is detecting credits across my library with a very high success rate.

In this video, you can see the detection process running and finding credits successfully. The only ones being skipped are items that already had credits markers detected previously, which is expected behavior.

This is a huge step forward, and I’m excited to keep improving it with more testing, optimizations, and additional features.

 

 

[yocker 1764]

47 minutes ago, EncryptedCity said:

I’m a little surprised you were able to reach that conclusion only 23 minutes after the post went up though, since the documentation and changes are fairly extensive. I’d be interested to know which parts you looked at that made it appear to be just a reskin.

Because it's not really hard to spot the similarities.
Many of the UI elementals are precisely the same. Even the tracer feature which i gave that name because i thought it would be a fun and original name to give it.


This is from EmbyCredits.
Plus many more examples.

As said, i don't mind. Just want clarification if it's EmbyCredits underneath.
 

[EncryptedCity 17]

Thanks for clarifying. I understand why some similarities stood out, especially with the UI elements and the tracer naming.

To answer your question: yes, it started from the EmbyCredits foundation, and I should have been clearer about that relationship. The goal was never to present it as if the original concept came from nowhere or to hide the connection. It was developed as an extension/evolution of the idea, while also remaining compatible with the original plugin.

A lot has changed since then. The OCR service, CPU/GPU OCR support, detection engines, processing logic, configuration system, and many other components have been redesigned, rebuilt, or expanded. The detection engines themselves were developed from scratch rather than simply being copied over.

The backend is also completely different in terms of architecture. It now uses dedicated Docker OCR services:

• CPU OCR Edition: ~1.89 GB container image

• GPU OCR Edition: ~8.34 GB container image

• Separate CPU and GPU endpoints

• GPU edition uses CUDA acceleration with PaddleOCR

• Dedicated OCR microservice architecture instead of relying only on the plugin process

It was also listed under Tools & Utilities because these services are designed to work with EmbyCredits as well. Users can use the OCR services with the existing plugin or use the expanded project alongside it. Both can be installed simultaneously.

The tracer naming similarity is a fair point — I should have picked a more unique name there to avoid confusion.

I appreciate you bringing it up. Hopefully this clarifies that this is not simply a reskin, but an expanded implementation built from the original concept with a significantly different backend and additional functionality.

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

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