Hi, > This is proof of why this intent of collecting command lines is totally > misleading. This will work *in certain cases*. What if the text page is > different? What if it's not run on a Raspberry Pi? (It won't work > anywhere else!). You need to understand both the input and the intent, > in order to understand a command line. (Kudos to Robin for > complementing his command line with that.)
I think this is a good point, the build configuration and the environment differences for each “working” command are going to make it hard to find any coherent pattern… I sort of had an idea, since I think I saw ML being brought up as a possibility, and used to find natural language patterns? I thought maybe if you offered a simple http service for people to check for common workflows and run them on a micro task server instance(?) (like lambda on aws or cloud f(x) on gcloud). If you do something like that then you’d be providing an incentive for people to provide working command lines (you have to make sure they agree to collection of commands for machine learning) plus you control the build configuration of the deployed function as a service endpoint. I can’t find it right now, but I remember seeing a site that did something like this, except it wasn’t based on any ffmpeg binary, and just sort of “parsed” the ubuntu man page and used the recognized options to parse the command, so it wasn’t very useful. But if you are able to build a good enough model, I think at least a few people would appreciate something like a web-based ffmpeg “terminal” with auto-complete suggestions, etc. And you could use information collected to further better your model. You could even make use of that data as an ffmpeg autocomplete rule set for popular shells, etc as well, Regards, Ted Park _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
