Create reference encodings manually or by using
python createVars.py [input] [fps] [maxmimum-durations]
Specify for [maxmimum-durations] a list for the maximum segment duration lengths, i.e. 4,6,8,10.
Extract the timestamps of the I-frame positions by using
python getTimeStamps.py [input m3u8] [fps] [sub]
Enable substraction of 1/4 frame duration by replacing [sub] with true. We use FFmpeg to extract the "best-effort-timestamps" of all I-frames that are at the beginning of a segment. To cope with floating point imprecision we subtract a small safety margin (1/4 frame duration). This ensures that the next frame after the adjusted timestamp is the frame which shall be encoded as I-frame. When using force key frame with this timestamp the next frame will be encoded as key frame.
Add the extracted I-frame timelines to the dictionary [Bunny/Elfuente/Meridia/TearsOfSteel]_t_sub in generateJobs.py. Ignore the dictionaries with the exact timestamps.
It should follow this style:
| Max Duration | Timeline |
|---|---|
| 4 | "0.0,3.98958,4.40625,8.40625,..." |
| 6 | "0.0,4.19791,10.19791,16.19791,..." |
| 8 | "0.0,4.19791,12.19791,16.28125,..." |
| 10 | "0.0,4.19791,14.19791,16.28125,..." |
Add the maximum segment duration for all variable-length encodings in generateJobs.py.
For the next step, you need to have the extracted average segment lengths of the variable-length encodings and round them to a precision of 0.5s. You can do this by first creating the variable-length encodings jobs python generateJobs.py and use the video docker to create the encodings and the video statistics files. Afterward, you can use
grep -oP '"durations_mean":\s(\d+\.\d+)' video_statistics.json
to extract the average segment length of the corresponding encoding. If you did this, you can remove the durations for variable lengths from generateJobs.py, otherwise, you will create those job files again in the next step.
Add the corresponding average lengths and the maximum lengths for the fixed lengths encodings in generateJobs.py. In the end, you can run
python generateJobs.py
to create all variable bitrate jobs.
Add the job files to the working queue of the docker container and start the encoding process.
Use
python extractCBRStats.py [DIR] [OUTPUT]
to extract needed bitrate information. The [DIR] input parameter should point to the result directory of the VBR Jobs.
The [OUTPUT] parameter should specify the output of the program as a JSON file, e.g. cbr_jobs.json.
Use
python generateJobs.py cbr_jobs.json
to generate all constant bitrate encoding job files.