Programmatic output?

[mjc]

It would be fabulous if there was an easy way to spit out json or some other kind of easily parsed output.

Is this something you'd take a PR for?

[alexheretic]

This does exist for sample-encode with the --stdout-format option. The idea of this is you can use this in a script to do custom stuff like a different search approach to crf-search.

It should be fairly doable to add similar --stdout-format support to crf-search. We'd need to flesh out the motivation, requirements/desired behaviour of this.

[mjc]

I'm building a web ui in my off time. I hope to open source it by the end of the year, I am tired of using shell scripts. I really dislike the closed source solutions that exist, and the open ones are too complicated for my use case.

It's basically a replacement for fd [...] -x ab-av1 auto-encode [...] that will pick either the fastest file to encode or the biggest size reduction first.

It would be very helpful for the programmatic mode for crf-search to have the input file and target vmaf on every possible line, at a minimum.

Here's the regexes I'm using atm, with named captures for bits I'm using. I can put more comments in about usage next time I have time outside work hours, if you need:

 @encoding_sample_regex ~r/
    # I plan on making a chart for these. input file would be helpful, as well as target vmaf.
    encoding\ssample\s
    (?<sample_num>\d+)\/             # Capture sample number
    (?<total_samples>\d+)\s          # Capture total samples
    crf\s
    (?<crf>\d+(\.\d+)?)              # Capture CRF value
  /x

  @simple_vmaf_regex ~r/
    ^-\scrf\s
    (?<crf>\d+(\.\d+)?)\s            # Capture CRF value
    VMAF\s
    (?<score>\d+\.\d+)\s             # Capture VMAF score
    \((?<percent>\d+)%\)             # Capture percentage
  /x

  @sample_regex ~r/
    sample\s
    (?<sample_num>\d+)\/             # Capture sample number
    (?<total_samples>\d+)\s          # Capture total samples
    crf\s
    (?<crf>\d+(\.\d+)?)\s            # Capture CRF value
    VMAF\s
    (?<score>\d+\.\d+)\s             # Capture VMAF score
    \((?<percent>\d+)%\)             # Capture percentage
  /x

  @eta_vmaf_regex ~r/
    # It would be helpful to have the input path here
    crf\s
    (?<crf>\d+(\.\d+)?)\s            # Capture CRF value
    VMAF\s
    (?<score>\d+\.\d+)\s             # Capture VMAF score
    predicted\svideo\sstream\ssize\s
    (?<size>\d+\.\d+)\s              # Capture size
    (?<unit>\w+)\s                   # Capture unit
    \((?<percent>\d+)%\)\s           # Capture percentage
    taking\s
    (?<time>\d+)\s                   # Capture time
    (?<time_unit>seconds|minutes|hours) # Capture time unit
  /x

  @vmaf_regex ~r/
    # currently I parse a bunch of stuff out of the filenames here.
    vmaf\s
    (?<file1>.+?)\s                  # Capture first file name
    vs\sreference\s
    (?<file2>.+)                     # Capture second file name
  /x

  @progress_regex ~r/
    \[.*\]\s
    (?<progress>\d+%)?,\s            # Currently I dont display this but I will.
    (?<fps>\d+\sfps)?,\s             # Currently I dont display this but I will.
    eta\s
    (?<eta>\d+\sseconds)             # Knowing ETA more often would be helpful even if its less accurate
  /x

  @success_line_regex ~r/
    # It would be helpful to have the path, target vmaf, percentage, vmaf score, time taken, and path in here.
    \[.*\]\s
    crf\s
    (?<crf>\d+(\.\d+)?)\s            # Capture CRF value from this one to know which CRF was selected.
    successful
  /x