[Suggestion] CF-Generated Thumbnails

I did write a script for this a while ago. Its still available on this post; https://hummy.tv/forum/threads/thumbnails-any-configuration.3241/page-7#post-43892

The parameters are at the top of the script, so it could easily be automated (or changed or ignored)
I've been running it as a cron job for years, and am very happy with the results.

The only issue is sometimes you get a failed (largely grey) thumbnail, but there doesn't seem any good way around this.
One of the options is to process in a slow mode for SD broadcasts when this happens (which usually works), but for HD it is probably too slow.
 
Nice, I had forgotten about that.

We have a much better framework for something similar to work in now, but maybe it's a starting point.
 
The webif libraries make this pretty easy to do as a one-off or cron-able task these days (although jxp's script deliberately re-creates any Humax-generated thumbnails too).

This would do for a one-off catch-up using a fixed 1-second offset.
Code:
#!/mod/bin/jimsh

source /mod/webif/lib/setup
require system.class ts.class

proc generate {ts} {
        set rfile [file rootname [$ts get file]]
        if {[file exists "$rfile.thm"]} return
        if {[$ts flag ODEncrypted]} return
        if {[$ts flag Radio]} return
        puts "Generating thumbnail for $rfile"
        $ts mkthm 1
}

ts iterate [lambda {ts} { generate $ts }]
 
We have a much better framework for something similar to work in now, but maybe it's a starting point.
I've knocked up a package to automatically generate thumbnails following decryption and to let you queue things for thumbnail generation via the browse screens. Is a fixed '1-second in' ok for you?
 
What would you adjust it to to get rid of the CA then so that everyone can have a share? Or is that just Milo in Catch 22?
 
I've knocked up a package to automatically generate thumbnails following decryption and to let you queue things for thumbnail generation via the browse screens. Is a fixed '1-second in' ok for you?
Fine by me.
 
I don't know yet!
Looking at my detectads.log files I see that the initial ad breaks that it finds and crops range from 3 seconds to over a minute with the majority seeming to be between 15 and 35 seconds.
This is for recordings primarily of ITV, Channel 4 and 5 with AR and no padding,
Since I don't run detectads on BBC channels I have no statistics of the average lead in there

I think it would be best to have the offset customisable via the settings page so that users can experiment to find the value that works best for them
 
I am pessimistic that any fixed value (or even AR marker + fixed value) will be much better than any other. It seems to me to make sense to replicate the Humax mechanism, and then use the WebIF thumbnail manipulation manually if it matters.

With the machine I use on padding, there will be a two-minute lead in as well as the difference between EPG scheduled start and actual start. The longer the offset is, the longer it will take to process.
 
Back
Top