Skip to content

Latest commit

 

History

History
243 lines (197 loc) · 12.9 KB

README.md

File metadata and controls

243 lines (197 loc) · 12.9 KB
Raw

About this Repo

This repo holds code and associated elements for automatically generating annotated WAVE mezzanine content from raw source material. The annotations in the resulting multimedia files are used by the WAVE test suite to programmatically identify content parameters and to verify various elements of performance. The WAVE Project is an industry interoperability effort for streaming internet video supported by over 60 companies and hosted by the Consumer Technology Association. For more information on the WAVE Project, the WAVE test suite, please see CTA.tech/WAVE or email [email protected].”

Contents

  1. Creating annotated WAVE mezzanine content
  2. Generating all annotated WAVE mezzanine content
  3. Experimental scripts

Creating annotated WAVE mezzanine content

The steps to create annotated WAVE mezzanine content have been combined into a single Python script mezzanine.py.

The mezzanine.py script does the following to the source content:

  • Adds timecode, frame number with configurable zero padding, frame rate, configurable label text and font:

    • Displayed in video.
    • Encoded in QR code displayed on each frame.
    • Option to configure QR code to alternate between either 2 or 4 positions.

  • Adds a low density bit pattern for each frame that signals:

    • Frame number.
    • Total frames.
    • Frame rate.
    • Horizontal resolution.
    • Vertical resolution.

    Note: The QR code is intended for black-box testing of devices using a camera to capture the video output, without requiring manufacturer involvement. The bit pattern is intended for white-box testing, where the video is processed directly within a device.

  • Adds border indications:

    • Red triangular indicators (assets/boundaries.png).
    • 2 pixel wide border around the edge of the video (outer 1px white, inner 1px black).

  • Integrates an A/V sync pattern at the top right of the video using modified synchronisation timing video test sequence generator scripts from DVB companion screen synchronisation timing accuracy measurement. The pattern is an irregular sequence of "beeps" and "flashes" that only repeats after a configurable duration (default 31 seconds).

  • Mixes source content audio with A/V sync beeps to allow for visual lip-sync confirmation.

  • Includes option to add a dark green (0x006400) color frame to indicate the start of the video, and a dark red (0x8B0000) frame to indicate the end of the video.

  • Optionally applies rudimentary tone-mapping to BT.709 SDR for each input file, to create SDR mezzanine streams using an HDR source.

  • Generates JSON metadata related to the mezzanine content, encoding, output file and source file. This includes the mezzanine release version, and the corresponding WAVE Test Content Format Specification version.

  • Video output configured to minimise lossy nature of transcoding and preserve properties of source content. The codec used depends on source content color space:

    • H.264/AVC for BT.709 or undefined
    • H.265/HEVC for BT.2020nc or other defined value

    Note: These codecs are used to ensure the properties of the content are signalled correctly. A more appropriate mezzanine video codec or raw YUV data may be used in future.

Requirements:

Using the provided requirements.txt, execute the following to install the Python dependencies: pip install -r requirements.txt

Associated assets:

  • boundaries.png, boundaries.xcf, red_triangle.xcf frame boundary markers overlay created in GIMP
  • Cousine-Regular.ttf monospaced font from the Cousine font family

Tools:

Mezzanine sources:

For example, you can execute the following command to produce an annotated mezzanine file:
py mezzanine.py --duration 60 --framerate 30 --label J1 --qr-positions 4 --resolution 1280x720 --seek 00:01:25 --start-end-indicators enabled --font assets/Cousine-Regular.ttf --window-len 6 --tonemap disabled --version 2 --spec-version 1 --metadata-only disabled source/tearsofsteel_4k.mov _mezzanine/tos_J1_1280x720@30_60.mp4

The full details of the available commands for the script can be found by executing py mezzanine.py -h

Generating all annotated WAVE mezzanine content

The metamezz.py Python script enables generation of multiple annotated mezzanine streams with a single command, using mezzanine.py to generate each annotated mezzanine file.

The requirements for metamezz.py are the same as for mezzanine.py.

At least one source file and an output file prefix pair must be specified, for example: py metamezz.py source1.mov source1_output_prefix

Additional parameters determine the specific annotated mezzanine files to generate (see further).

WAVE annotated mezzanine filenames use the following template: <prefix>_<label>_<WxH>@<fps>_<duration>.mp4

Where:

  • prefix is a short string based on the source content used, e.g. "tos" for Tears of Steel.
  • label is a short string used to identify a particular piece of test content, e.g. "A1", "L1", "L2".
  • W and H are the horizontal and vertical resolutions in pixels.
  • fps is the frame rate.
  • duration is the duration of the annotated mezzanine content, in seconds.

For example: croatia_O2_3840x2160@50_60.mp4 or tos_L1_1920x1080@60_60.mp4

To generate a complete set of WAVE mezzanine content you can execute the following 2 commands:

py metamezz.py "source\tearsofsteel_4k.mov" _mezzanine\tos "source\tearsofsteel_4k.mov" _mezzanine\tos "source\DVB_PQ10_VandV.mov" _mezzanine\croatia -rjf rjf\resolutions_15_30_60_fractional.json rjf\resolutions_15_30_60_non-fractional.json rjf\resolutions_12.5_25_50.json --tonemap disabled disabled enabled --spec-version 1 --version 2

py metamezz.py "source\tearsofsteel_4k.mov" _mezzanine\splice_main_tos "source\tearsofsteel_4k.mov" _mezzanine\splice_main_tos "source\DVB_PQ10_VandV.mov" _mezzanine\splice_main_croatia "source\Big Buck Bunny trailer_1080p30.mp4" _mezzanine\splice_ad_bbb "source\Big Buck Bunny trailer_1080p30.mp4" _mezzanine\splice_ad_bbb "source\Big Buck Bunny trailer_1080p.mov" _mezzanine\splice_ad_bbb -rjf rjf\splice_main_resolutions_30_fractional.json rjf\splice_main_resolutions_30_non-fractional.json rjf\splice_main_resolutions_25.json rjf\splice_ad_resolutions_30_fractional.json rjf\splice_ad_resolutions_30_non-fractional.json rjf\splice_ad_resolutions_25.json --tonemap disabled disabled enabled disabled disabled disabled --spec-version 1 --version 2

This assumes the source files are located in the source folder, and the JSON files in the rjf folder. The annotated mezzanine files generated are saved to the _mezzanine folder, with the prefixes tos and croatia for the main mezzanine content, and with the prefixes splice_main_tos, splice_main_croatia and splice_ad_bbb for the splicing mezzanine content. Tone-mapping is enabled for the croatia content, as the source is HDR, but the desired output for the mezzanine content is SDR. The mezzanine release version is set to 2, and the corresponding WAVE Test Content Format Specification version is set to 1.

The following set of JSON files is used to generate all WAVE annotated mezzanine content. These JSON files are passed to metamezz.py using the -rjf parameter.

  • Main annotated mezzanine content (60 second duration):
    • resolutions_12.5_25_50.json -- 12.5/25/50 fps
    • resolutions_15_30_60_fractional.json -- 14.985/29.97/59.94 fps
    • resolutions_15_30_60_non-fractional.json -- 15/30/60 fps
  • Shorter main annotated mezzanine content for splicing tests (10 second duration):
    • splice_main_resolutions_25.json -- 12.5/25/50 fps
    • splice_main_resolutions_30_fractional.json -- 14.985/29.97/59.94 fps
    • splice_main_resolutions_30_non-fractional.json -- 15/30/60 fps
  • "Ad" annotated mezzanine content for splicing tests (duration depends on frame rate):
    • splice_ad_resolutions_25.json -- 12.5/25/50 fps (5.76 second duration)
    • splice_ad_resolutions_30_fractional.json -- 14.985/29.97/59.94 fps (21.255 second duration)
    • splice_ad_resolutions_30_non-fractional.json -- 15/30/60 fps (6.4 second duration)

The JSON structure used is: { "WIDTHxHEIGHT" : [ [framerate (str), duration in seconds (float), starting position in source (str, HH:MM:SS), label (str), number of variants (int), add second audio track (bool)], [...] ]}

Multiple combinations of frame rate, duration, start position, label and variants (with/without second audio track) can be defined for each resolution.

Details of the available commands for the script can be obtained by executing py metamezz.py -h

Here is a brief description of the parameters:

  • -m enabled || disabled disables mezzanine generation and only (re)generates JSON metadata using existing mezzanine files. The source and output mezzanine files must both be present at the paths provided.
  • -r <string_containing_JSON> or -rjf <path_to_JSON_file> that provide JSON defining the following properties of the generated mezzanine content:
    • Resolution (width x height).
    • Frame rate (string, fractional rates must be specified as division operations, e.g. 30000/1001).
    • Duration (in seconds).
    • Starting position in source content (string with timestamp using the HH:MM:SS notation).
    • Label (one or more characters, e.g. 'A', followed by a number, appended automatically for each variant starting from 1). Labels are used to identify the different annotated mezzanine streams. They are displayed in the video, encoded in the QR codes displayed in the video, and included in the output filename. The character(s) identify the particular combination of {resolution, frame rate, duration}. The following number identifies the variant (see below). E.g. 'A1', ... 'ÁN', 'B1', ..., 'BN'.
    • The number of variants to create for each combination of {resolution, frame rate, duration, starting position, label}. A variant is a duplicate of the same mezzanine content with a different label (same letter, different number). Specifying N variants for a combination with label 'A' would result in mezzanine streams with labels 'A1' to 'AN'.
    • Flag (boolean) that determines whether to generate a second audio track for streams from a particular combination of {resolution, frame rate, duration, starting position, label, number of variants}.
  • --tonemap [enabled || disabled, ...] enables or disables rudimentary tone-mapping to BT.709 SDR for each input file, to create SDR mezzanine streams using an HDR source. Provide one value, and it will apply to all input source files. Alternatively, provide one value per input source file, separated by a space.
  • --test 1 || True is a flag indicating a test run, which will parse the parameters and list the streams to generate, but won't actually generate the streams.

The metamezz.py script uses the following parameter defaults that can only be modified in the script, as they are not expected to be changed often, for consistency reasons:

  • The script expects the presence of mezzanine.py in the same folder.
  • The default border indicators are used and boundaries.png is expected to be in the assets folder.
  • The font is set to Cousine-Regular.ttf and is expected to be in the assets folder.
  • 4 QR code positions are used.
  • Start and end indicators are enabled.
  • The irregular AV sync pattern is set to repeat after 63 seconds.

Experimental scripts

Adding a second audio track to a mezzanine stream

The add_second_audio_track.py Python script uses the pyttsx3 library to create an audio file with the spoken audio "English". This is mixed with the original audio of a mezzanine file, repeating the spoken audio every 15 seconds.

The script makes a copy of a mezzanine file, incorporating the new audio track as a second audio track. The first audio track remains the same as in the original mezzanine file.

The output file naming convention is as follows: <mezzanine_stream_name>_2ndAudio[English].<mezzanine_stream_file_extension>

Usage example: add_second_audio_track.py tos_A1_480x270@30_60.mp4 creates tos_A1_480x270@30_60_2ndAudio[English].mp4

Additional requirements: