opentimelineio / otio-youtube-example Goto Github PK

See this comment

We would like to explain OTIO in five levels of complexity. Currently, the documentation covers the first two levels only.

• Level 1: The high level purpose of OTIO
• Level 2: OTIO is a file format, an API, and a collection of adaptors and plugins.
• Level 3: The technology stack of the OTIO codebase
• Level 4: Important design patterns / architectural components
• Level 5: Walkthrough of the codebase

To test your changes, follow the instructions on installing MkDocs.

How to test:

• Verify that an .otio gets generated on disk.

• Verify that the Timeline has the correct number of clips.
• Verify that available_range is set correctly.

Currently, the regular expressions that we use to parse time stamps assume that they will have the following format:

for example,

1:47 Introduction to hash maps

But there are other timestamp formats such as in this video. Can you improve the parsing logic to work with the format in that video? Also, are there other timestamp formats featured in other videos that we're not aware of?

We would like to illustrate a plugin within the otio-youtube-example. Plugins are an interesting concept within OTIO which may be confusing to understand for newbies. If we can show a concrete example of a plugin within our repo, we could help to clarify what a plugin is, how it is used, and how to create one.

Here is some feedback from Joshua Minor on a possible direction we can explore to create a plugin:

"... make a Media Linker plugin which looks at individual clips in an OTIO, and if they point to a YouTube URL, it could pull metadata & markers into the clip from YouTube."

https://opentimelineio.readthedocs.io/en/latest/tutorials/write-a-media-linker.html

See reneicke's comment

Adding a linter can uncover unexpected problems and align the style/formatting to a standard.

See this comment

See ssteinbach's comment

See this comment

Create something similar to this (but more user-friendly) that documents our Youtube/OTIO example.

See jminor's comment

See jminor's comment

See reneicke's comment and ssteinbach's comment

Also see this comment and this comment

Right now, the unit tests require youtube-dl to finish downloading a complete video from Youtube. Ideally, we can stub some of this data (mp4 file and the description file) so that the tests can still run even if youtube-dl is not functioning.

Based on jminor's feedback:

"I always worry about timezones accidentally causing calculations like this to vary."

Also see Reinecke's comment about using timedelta.

A youtube video description could have timestamps in a variety of formats:

• 01:34:21 chapter title
• 1:34:21 chapter title
• 34:21 chapter title
• 4:21 chapter title
• (4:21) chapter title
• [4:21] chapter title

We should have a test cases to confirm that each of these formats is properly handled.

See ssteinbach's comment

Somehow, even though the example was crashing with an error (undefined variables), the tests still passed. See example screenshot.

.

Ideally, the tests should have flagged some kind of failing test case.

See jminor's comment

Also see reneicke's comment

Create a single clip on the timeline with a media_reference to the original youtube video file

How to test:

• Verify that the clip has the correct media_reference.
• Verify that the Timeline has the correct number of clips.
• Verify that available_range is set correctly.

See jminor's comment and reinecke's comment

Confirm that the markers have the correct timestamp and title.

• Marker's time-stamp should have the correct frame number
• The title should match whatever is listed next to the time-stamp in the youtube description

Parse the video description in python and create a Timeline object with Markers to represent the chapters

Add a check before the main video description parsing logic which confirms that the video description actually is text and that it contains time stamps.

How to test:

• Verify that the Timeline object has the correct of Markers on it.
• Verify that each marker has the correct text on it.

See jminor's comment and reinecke's comment

Right now, most of the logic for downloading the video from youtube and importing the data into the OTIO Timeline is in one long block of code. It would be better to refactor this code to use more helper functions.

See this comment

See ssteinbach's comment

Create a python script that takes as an input argument the url for a youtube video. The script should use youtube-dl to download the video file and the video description.

How to test:

• Verify that an mp4 and txt file is created on disk.

Ideally, the command to run the example could look like the following:

python3 youtube_chapters_demo.py pvkTC2xIbeY --skip-video-download