todokku / video-generator

A workflow to generate videos using Amazon Polly and YouTube

video-generator

A workflow to generate videos using Amazon Polly and Translate. This script serves to complement edx-generator by processing videos stored in the input folder of an unit.

The aim is:

• to shorten a video creation timeframe by employing modern text-to-speech services by AWS Polly

• facilitate creation of educational videos from common instructional formats like powerpoint slides

• increase reach of videos by employing automated translation services

Requirements

Listed below are essential packages used in the script

To Download

• unoconv
  • Download as zip. unoconv file (without ext) required in subsequent installation steps

Manual Installations

• Image Magick
  • 7.0.9-5-Q16 (breaks with latest: Q8)
  • used by moviepy in the video creation process
  • conversion of .pdf to .png
• Libre Office
  • used by unoconv to convert pptx to pdf
  • used by unoconv to convert pptx to xml
• GhostScript
  • required by unoconv to facilitate conversion to pdf

pip packages

• boto3
  • AWS python SDK
  • Services used: Polly, Translate, and S3
• moviepy
  • version 1.0.0
  • WINDOWS USERS: provide the path to the ImageMagick binary called magick.exe in PATH/TO/PYTHON/lib/site-packages/moviepy/config_defaults.py
• bs4
• lxml

Input

Settings

__CONSTS__.py

1. Course Settings
   • Course Path should be set in a __RUN__.bat file. Refer to Execute Section #6
   • Course path: Relative Course Path (from root folder) where units will be checked. Script iterates 3 levels deep to get to the units from the Course Path

     `-- input
         `-- Course (Starts here)
             |-- Section
             |   |-- Subsection
             |   |         `-- Unit (Target folder)
             |   `-- Subsection
             |       `-- Unit (Target folder)
             `-- Section
                 `-- Subsection
                     `-- Unit (Target folder)
     

   • Expects a __SETTINGS__.py sibling folder

     `-- input
         |-- Course
         |
         `-- __SETTINGS__.py
     

     • File should include the following CONSTANTS:
       • S3_BUCKET
       • S3_MOOC_FOLDER
       • S3_VIDEOS_FOLDER
       • LANGUAGES
     • Videos will be uploaded onto: S3_BUCKET\S3_MOOC_FOLDER\S3_VIDEOS_FOLDER
2. Video settings
   • Resolution: (width, height) in px
   • Title Period: Period in which the title is shown (in seconds)
   • Font: Font to be used for video title
   • Title ideal length: length in characters. Title will be split based on specified length
   • Font size: Font size to use for title
3. Audio settings
   • Languages: Languages which will be used to generate the videos
   • Voices: Based on list available on AWS Polly Documentation
     • Take note whether neural voice is available for the voice
     • voice key is used for AWS Translate

Course Units

1. Video Settings within SRT/PPTX notes
   • Must be located in the first sequence/slide
   • Follows string JSON format:
     • video_file_name: file name for generated video
     • display_name: text for generated title
     • voice: voice to use for each language; Mapped to index of array in __CONSTS__.py

     {
         "video_file_name": "FILE_NAME",
         "display_name": "VIDEO TITLE",
         "voice": "1"
     }
     

   • {"display_name":"break title"} may be used to insert a title within a video
2. MP4 file with srt
   • srt sequence should start at index 1 and be appended with _language
   • for optimum results, ensure srt sequences are full sentences and use the start time to determine points where speech should start

     `-- input
         `-- Course
             `-- Section
                 `-- Subsection
                     `-- Unit
                         |-- mp4_file_name_en.srt
                         |-- mp4_file_name.mp4
                         `-- mp4_file_name.yaml               
     

3. pptx file
   • transcripts are to be written in presenter notes
   • for optimum results, ensure presenter notes are in full sentences

     `-- input
         `-- Course
             `-- Section
                 `-- Subsection
                     `-- Unit
                         |-- pptx_file_name.pptx
                         `-- pptx_file_name.yaml
     

4. Accompanying language-specific srt files
   • script will use provided srt files when available
   • follows the same naming convention as shown above
   • for optimum results, ensure srt sequences are full sentences and use the start time to determine points where speech should start
   • note that translated srt files will not be replaced with a new translation when the core pptx or mp4 file, or _en.srt file has been modified. Text-to-speech is based on said srt file and may not be up to date with the new modifications. To generate a new non-english srt, delete the language srt file.

     `-- input
         `-- Course
             `-- Section
                 `-- Subsection
                     `-- Unit
                         |-- pptx_file_name_fr.srt <--
                         |-- pptx_file_name_zh.srt <--
                         |-- pptx_file_name.pptx
                         `-- pptx_file_name.yaml
     

Output

1. videos.json file
   • This file serves as a log to keep track of last modified time for all the pptx, mp4, and their srt files.
   • deletion of videos.json or renaming of files will result in regeneration of the videos

   `-- input
       `-- Course
           `-- Section
               `-- Subsection
                   `-- Unit
                       `-- videos.json <--
   

2. language specific sub files
   • The sequences are split at punctuation marks and shortened for a smoother performance
   • They should not be reused for audio generation as the sentences will be incomplete

   `-- input
       `-- Course
           `-- Section
               `-- Subsection
                   `-- Unit
                       |-- example.mp4
                       |-- example_en.srt
                       `-- example_sub_en.srt <--
   

3. translated srt files
   • The sequences are translated as-is from the original en srt file or pptx notes.
   • This file may be edited to regenerate the video of its language

   `-- input
       `-- Course
           `-- Section
               `-- Subsection
                   `-- Unit
                       |-- example.mp4
                       |-- example_en.srt
                       `-- example_zh.srt <--
   

Example

w1 > s1 > u1

Execute

1. Rename __AWS__.template.py to __AWS__.py
   • Include AWS Access Key and Secret Access Key in renamed file
2. Add to Environment Variable LIBRE_OFFICE_PROGRAM with value path\to\LibreOffice\program
3. Copy unoconv (no ext) into path\to\LibreOffice\program\
4. Add to Environment Variable Path with value path\to\ImageMagick-XXXX
5. Add to Environment Variable IMAGEMAGICK_BINARY with value path\to\ImageMagick-XXXX\magick.exe
6. Rename __RUN__.template.bat to __RUN__.bat
   • set path to Course folder
7. Execute

   .\__RUN__.bat