We are taking a peek at this show based on everyone’s feedback!
Now that you’re here Tablo, Why does it work on a show (Jeopardy, TMZ…etc) one day and the next it doesn’t work at all with no massage or why it did not Commercial Skip?
I think it got worse after the last update, that said it was going to make it better.
I would have saved the episode but your Poll didn’t include a category for our watching habits - Delete recording at Warp speed.
It’s possible that these recordings have a different error type that isn’t yet configured for notification. This particular error happens if there are more or fewer ad pods than expected for that length of program, or if those ads are much longer/much shorter than expected.
This error type will be categorized and given a notification during some upcoming updates.
That’s a SWAG.
This show has had the consistent advertising points and lengths for years. Yesterday all the ads except the last ad were detected. Today the last ad was there, it was the skip for the middle of a game…
Your points are valid.
Just some fun with your analogy, there’s different types, grades and styles of cheeses. Some misunderstand the plastic wrapped pasteurized cheese food product as “cheese”. Your ordered a cheese burger… and got the solidified slime 
.
I find that CS seems to work great most of the time with main TV channels. Not so good with most of the subchannels, but even some them are OK. Another observation: when I improved my internet from 5 down to 60 down the CS worked much better. And, when I replaced my 2014 model quad with a 2021 quad things got even better. Still not perfect but good enough for $20. Just my 2¢
The recording that it did nothing on was the same recording that Commercial Skip worked fine the day before. So how could it have a different error type when it was the same program with the same Commercials? This same thing happens on other shows too. The program was not even attempted to be Commercial Skipped from what I see. How many times are your servers down?
Is Commercial Skip a Tablo Program or a third party program?
@kpatter00 @zippy The engineering team is going to be taking a closer look at the Price Is Right to see if they can tweak the algorithm and get things working a bit better.
Would you be willing to have them look specifically at the uploaded commercial skip samples from your Tablo assuming you record the show every day?
If so, send me a DM with the email connected to your Tablo subscription account or your Tablo’s serial number.
Sure you can look at our recordings.
Letme know if youneed anything else.
Got it and removed it for security reasons. 
Thanks for your help!
@kpatter00 In yesterday or today’s episodes, did you notice the same behavior? Either missed or an early skip or both?
I opened a ticket and included my S/N.
I would love to tell you about today’s episode but it didn’t generate commercial skip. I would also love to tell you why but the last firmware update seems to have broken the return error messages.
We aren’t home right now and will check it out when we get home
1 Like
Nothing was ‘broken’ but there are some newer errors that aren’t yet getting passed back to the user.
Those errors are when there are more or fewer ad breaks than expected for that length/type of recording and that’s exactly what happened with your Price Is Right today… More ads than expected.
About 20% of my recordings fail CS processing. And the only one(s) that had a message indicating why was when the tablo AWS AI processors were down a week or so ago. Otherwise there is no message. Such as last nights James Corden Late Late Show.
And if you are using AI with Machine Learning and Digital Signal Processing what does “more ads then expected” even mean. You are either processing the digital signal as it is in the episode or you are trying to apply some template to a show.
1 Like
Tablo, So when you say “More ads than expected” does that mean that the show has to be the same every time or Commercial Skip wont even try?
There is a difference between not processing some of the commercials in a show properly and not processing any commercials in a show.
The big 5 networks(ABC,NBC,CBS,FOX,CW) distribute TV shows to their affiliates. So what tablo seemed to be saying is that they know that some of these TV shows can’t be processed by CS at all. Yet there doesn’t seem to be a list of these network shows.
Well if they know what cant be processed that should be an easy list to show us. My problem is a show that CS’s perfectly and then the next day not at all with no reason and I know the server is not down because other shows do CS.
We do provide a list of programs that are 100% ineligible here:
This is different though.
If you’ve ever watched one of those ‘gold mining’ shows on the History or Discovery channel you’ve probably seen how they use progressively finer filters to get to the good stuff which then gets processed.
Our commercial skip server uses a similar method. We do very basic checks of the recording before processing it to ensure that what comes out the other end is what we (and you) expect to see.
First, we automatically eliminate shows from networks with no commercials (like PBS) and shows we know will perform poorly. (The list above.)
Then we do additional checks for other red flags like super short recordings which are rejected.
(See Automatic Commercial Skip Error Messages – Tablo)
As mentioned above, one of the checks we do is to determine if the recording has many more or much fewer ads than expected for that type and length of program.
For example, there’s an average amount of ad breaks expected for a 1 hour gameshow like the Price Is Right.
But let’s say one day part of the program is pre-empted by a press conference so there are 3 fewer ad breaks than we expect.
In order for the AI & DSP tools we use to identify and mark the commercial breaks isn’t confused by this odd outlier, potentially spitting out a recording with erratic skips, we choose instead to not completely process that recording.
Automatic Commercial Skip isn’t 100% perfect. Even if we used real humans instead of computers it wouldn’t be 100% perfect.
But we’re always working to improve the tool’s accuracy which is why having you guys flag results where there’s a consistent error on a specific show is super helpful.
How big a difference is there between an episode with no CS and one with incorrect CS.
For me, James Corden seems to rarely generate CS.
Last nights Jimmy Fallon did generate CS. But the first CS start 40 seconds before the end of Fallon’s opening monologue (something about cupcakes - I think). And then during the Dave Chappelle interview there was the 20+ second phantom CS. Both of these cause a user to rewind to see what happened.
What luck I usually delete recordings once i watch them but I guess I hadn’t finished this one. So CS went off the rails at 11 minute and 18 minute mark.