https://github.com/hekmon/pgs-ai-ocr
Extract PGS subtitles and extract text using external Vision Language models.
https://github.com/hekmon/pgs-ai-ocr
ai ocr ocr-recognition openai-api pgs pgs-subtitles srt srt-subtitles subtitles vision-language-model
Last synced: 11 months ago
JSON representation
Extract PGS subtitles and extract text using external Vision Language models.
- Host: GitHub
- URL: https://github.com/hekmon/pgs-ai-ocr
- Owner: hekmon
- License: mit
- Created: 2025-05-12T07:31:02.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-05-30T15:20:02.000Z (about 1 year ago)
- Last Synced: 2025-06-26T21:44:16.295Z (12 months ago)
- Topics: ai, ocr, ocr-recognition, openai-api, pgs, pgs-subtitles, srt, srt-subtitles, subtitles, vision-language-model
- Language: Go
- Homepage:
- Size: 53.7 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# PGS AI OCR
Extract PGS subtitles and recover text using an external (with OpenAI API) Vision Language model to create an output SRT file.
## Usage
Once the subtitles are written to the `.srt` file, I recommend you to open it with [Subtitle Edit](https://github.com/SubtitleEdit/subtitleedit) to:
- Check for the model output (if you did not use the `-debug` flag)
- Optimize it with the "Tools > Fix common errors" utility
```raw
Usage of ./pgs-ai-ocr:
-debug
Print each entry to stdout during the process
-input string
PGS file to parse (.sup)
-italic
Instruct the model to detect italic text. Not all models manage to do that properly.
-model string
AI model to use for OCR. Must be a Vision Language model. (default "o1-mini")
-output string
Output subtitle to create (.srt subtitle)
-timeout duration
Timeout for the OpenAI API requests (default 30s)
-version
show program version
```
### Simple (OpenAI)
#### Linux/MacOS
```bash
export OPENAI_API_KEY="your_openai_api_key_here"
# you can validate with the following cmd: echo $OPENAI_API_KEY
./pgs-ai-ocr -input /path/to/input/pgs/subtitle/file.sup -output /path/to/output/subtitle/file.srt -debug
```
#### Windows
Using the command line (`cmd.exe`):
```bat
set OPENAI_API_KEY=your_openai_api_key_here
:: you can validate with the following cmd: echo %OPENAI_API_KEY%
.\pgs-ai-ocr.exe -italic -input C:\path\to\input\pgs\subtitle\file.sup -output C:\path\to\output\subtitle\file.srt -debug
```
### Advanced (self-hosted)
For local inference [Qwen2.5-VL 7B](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) is recommended for best results (even if it can not handle properly italic text).
#### Linux/MacOS
```bash
# using vLLM
./pgs-ai-ocr -baseurl http://127.0.0.1:8000/v1 -model "Qwen/Qwen2.5-VL-7B-Instruct" -input /path/to/input/pgs/subtitle/file.sup -output /path/to/output/subtitle/file.srt -debug
```
#### Windows
Using the command line (`cmd.exe`):
```bat
:: using vLLM
.\pgs-ai-ocr.exe -baseurl http://127.0.0.1:8000/v1 -model "Qwen/Qwen2.5-VL-7B-Instruct" -input C:\path\to\input\pgs\subtitle\file.sup -output C:\path\to\output\subtitle\file.srt -debug
```
## Thanks
[@mbiamont](https://github.com/mbiamont) and its [go-pgs-parser](https://github.com/mbiamont/go-pgs-parser) library