Nowadays I just ask ChatGPT to give an “ffmpeg” terminal command
Works much faster than any app and I can customize it
Works much faster than any app and I can customize it
The handbrake GUI has a drop down list of encoding presets. I find that as an amateur, selecting one of those presets is the best way to make a file smaller.
> I find that as an amateur, selecting one of those presets is the best way to make a file smaller.
These presets are not helpful when you want to try to make something specific like "I want to make this file fit in 3GB" which is something that an amateur typically wants to achieve.
It's not even that hard to actually do, which makes me wonder why Handbrake has never implemented this kind of things.
After trial and error, I found the original file size to be preserved and not to correspond to the presumed smaller output of say a lower resolution output.
It would also work with a one-pass encode, but setting the bitrate that way risks wasting space in simple parts of the video and degrading quality in complex parts. Two-pass encoding takes longer but distributes bandwidth better.
The problem is that "retaining X quality" is extremely vague and subjective. Also, a lot of people conflate resolution with quality (see YIFY Torrents). While yes, you need enough pixels to have reasonable quality, if the encoder set CRF to 50, 8K video won't save you.
There's very little that is user-friendly in Handbrake.
Much easier to google / llm for the right ffmpeg arguments
hahahahha
I’ve used the CLI tool for a lot of my media library and it really isn’t that hard, there are even wrapper scripts in GitHub that simplify the whole thing and are easier than handbrake
An app can be very user friendly, but does not give the user the power they didnt know they wanted. This is what handbrake is - you get a nice GUI, you get to choose from a list of presets (unless you know what you're doing and customize it). Then you click go - you can even queue up more files.
Someone who is using handbrake is not going to learn the command line. How would they know how to queue up ffmpeg? Don't say batch scripts, because that's not something a normal user would use.
It seems like presets still exist.
https://handbrake.fr/docs/en/latest/technical/official-prese...
I've tried fiddling around with when the subtitles should start, but because the subtitles end up stretched out so much, that makes all the later subtitles wrong. And I'm typically ripping a bunch of TV shows, so I want to do as little manual work as possible for each episode, because that all adds up quickly.
Do you know if I'm missing a library or a setting somewhere? Maybe I just need to try again with the latest version and see if it's been fixed over time.
I tried using the libraries from MakeMKV but found it easier to just two step the process.
If you care enough about subtitles and are converting for future reference then it's worth using SubtitleEdit to correct | align | correct case | spell check | generate translations | etc and then merge final (video + audio) tracks from HB with SubtitleTracks using (say) MKV Toolnix.
https://nikse.dk/SubtitleEdit/
These are tools that can be streamlined and batched (with some degree of learning curve).
> so I want to do as little manual work as possible for each episode
I typically transcode Audio+Video, seperate out subtitles automatically along with "most common least destructive" touchups scripted and then watch.
You can achieve this by dropping input file in a watched folder and having the results plopped out in a "to be viewed" folder.
Most of the time everything is A-OK .. when the syncing is out I correct it via SubtitleEdit and continue watching.
You can save Episode.mkv and Episode.srt together, or batch bind the srt into the mkv as you store OR you can go to town with multiple subtitles if you're a media meta data nerd.
I rely on gaupol for subtitle editing.
I use both. One thing I prefer about ffmpeg is that I prefer it when I'm being a control freak.
Basically, if I'm encoding a lot of videos, I will sometimes fail to notice that some box has been ticked in Handbrake, that will screw up my encode. In particular, Handbrake always defaults to resizing my videos, 100% of the time, and I always have to turn that off manually. If I fail to do that, it will resize my videos and I'll end up wasting an hour doing encodes, or 3-4 hours if I'm using CPU instead of GPU.
Which implies you actually understand all the ffmpeg opts and how they interact.
I very much doubt trail an error, or reading manpages is "mich faster" than picking a preset, ticking a box or dragging a slider in handbrake.
As long as you have this behavior for non-critical code it's just a little sad because you are delegating something you could easily have learnt. For hype sake I guess.
But if you do use this technique in a work related context you are just going to produce average code that you won't be able to debug when something break...
This is the only case I would ever use a text generator. If you cannot understand it, you cannot trust the output and you cannot learn it in case of doubt.
This is why it is so great for grammar and protocol, but very problematic for actual research questions.
[0]: https://github.com/oyvindln/vhs-decode/wiki/CX-Cards#what-is...
Feed ChatGPT's output to a question, back as input to it, say it was from a human (1, amateur, 2, expert) and for each case, for that question, ask it if it is correct.
A ChatTuring (pronounced chattering) Test.
But now I'm seeking a --chatgpt option similar to --help so I can navigate any man page.
If you have true experience in some field try to ask ChatGPT some questions and you will be shocked what nonsense it suggests, put in very nice words.
From a random blog post that a search machine brings up you can often get some clues whether the author has good understanding or just wrote down a random finding they had during trial and error. And that finding is still more on the side of error than doing it correctly.
In ChatGPT answers I don't see such hints.