Skip to content

Latest commit

 

History

History
150 lines (127 loc) · 5.06 KB

README.md

File metadata and controls

150 lines (127 loc) · 5.06 KB
Raw

Introduction

this project is about transcribing the english video files
and creating english subtitile and the translating and
adding other subtitles in other languages as subtitle to the video file.
the main Two repositories:

the supported languages are:

  • English
  • Spanish
  • French
  • German
  • Russian
  • Chinese
  • Japanese
  • Persian

image

Installation

1.run the command below:
pip install -r requirements.txt
for installing the requirements

2.run the command below:
python3 argos_setup.py
for installing the offline argos_translate packages
(note: either this or you can install the packages online) the packages that are used originally are as following:

  • translate-en_fa-1_5.argosmodel
  • translate-en_fr-1_0.argosmodel
  • translate-en_es-1_0.argosmodel
  • translate-en_zh-1_7.argosmodel
  • translate-en_de-1_0.argosmodel
  • translate-en_ru-1_7.argosmodel
  • translate-en_ja-1_1.argosmodel

and their link are as follows:

links to all packages:

After Downloading Models

do consider that file model names could change based to there versioning, if you got diffrent version's from above either you need to change
the argos_setup.py file or install them manually. you need to put the files inside the argos_packages_offline directory like the following structure:

argos_packages_offline/
├── en_ar.argosmodel
├── en_es.argosmodel
├── en_fr.argosmodel
├── translate-en_de-1_5.argosmodel
├── translate-en_fa-1_5.argosmodel
├── translate-en_ja-1_1.argosmodel
├── translate-en_ru-1_7.argosmodel
└── translate-en_zh-1_1.argosmodel

OpenAI Whisper

after installation the one of the whisper models should
be downloaded(multiple models could exist in ./openai/ directory)
in the following directory example:

openai/
├── whisper-base
│   ├── added_tokens.json
│   ├── config.json
│   ├── flax_model.msgpack
│   ├── generation_config.json
│   ├── merges.txt
│   ├── normalizer.json
│   ├── preprocessor_config.json
│   ├── pytorch_model.bin
│   ├── README.md
│   ├── special_tokens_map.json
│   ├── tf_model.h5
│   ├── tokenizer_config.json
│   ├── tokenizer.json
│   └── vocab.json
└── whisper-small
├── added_tokens.json
├── config.json
├── flax_model.msgpack
├── generation_config.json
├── merges.txt
├── normalizer.json
├── preprocessor_config.json
├── pytorch_model.bin
├── README.md
├── special_tokens_map.json
├── tf_model.h5
├── tokenizer_config.json
├── tokenizer.json
└── vocab.json

tiny whisper model is the smallest pretrained model.
you can download other ones and put them inside ./openai/
folder if you one to use another one, and also
select the model in model_select function in functions.py
and the put it in the line after model_select function as follow:

	#i.e if you downloaded the small model  
	def model_select(model_name):  
		.  
		.  
		.  
	 	if(model_name == 'small'):  
        		with open(log_file, "a") as log:  
            		log.write("using small model\n")  
        		return   pipeline("automatic-speech-recognition", model="openai/whisper-small")  

`pipe = model_select(model_name='small')`

i.e the small model link is https://huggingface.co/openai/whisper-small

cmd_test.py

if you want to use cmd_test.py , after downloading and
setting up the models, you need to rename your video file to video.mkv
and put it in cmd_test.py directory.
(cmd_test.py is for quick test)

Docker

for using Docker, build and run commands should run after the argos_translate and whisper models have been downloaded in their directories. command sample if you want to build and run docker:
docker build -t argos_whisper:latest
docker run --network="host" argos_whisper:latest
and the open: http://127.0.0.1:5000