Functionality: Define a set of cards that belong together. The time for review is calculated by Anki independently for each card just like it is now. What is different is how it is presented to the user. The whole string will be presented, in the predefined order for this string, but the string will stop for user interaction only for the cards that need a review. The delay between the cards that do not need user response could be also predefined or at the limit it could be 0.

What I have in mind is this. You select a scene from your favorite TV show, drama, anime, maybe also podcast? and it doesn't matter how long (1 min, 1 h, it's your time) You split it in cards (I'm thinking subs2srs now). You define the whole scene as a string in Anki. And then, every time you review, the whole scene is played by Anki with stops only at the cards that need to be reviewed by you at that time.

In this way you will have more context to your cards, and I think it will be more fun (it would be almost like looking at the show, with some stops for clarification). The downside is that it requires more time but maybe the benefit is greater than the loss here.

I know that the whole string of cards idea is not new but maybe it was not presented in this way.
I know that nuke is fond of the idea of dissecting whole dramas and icecream keeps only specific parts with context. Maybe this thing here would meet both requirements?

I would like to know what do you think about this? Is this making any sense to you?

Mmm. Nobody. I guess I have my answer.

Anyway in case somebody will read this, I am still thinking about it. The more I think, the more I realize it would be easier to attach an SRS engine to a player (maybe Mplayer). In this case you would not need to split the video into cards, just to stop it at specific times for the actual review to take place.

Anybody else interested?

I'm not sure exactly what exactly you want to do, but I think this might work:
Use subs2srs. For the cards you want to "string" together, just get the line with the word you want to remember. Set the offset from start to finish to fit the entire conversation. it might be kind of annoying to do it for every line in the conversation, or at least every word you want in the conversation, but wouldn't it get it done?

OR, it might be an option to use the "context" option to select however many lines before and after you want a card to encompass. That might work as well...

I think that might be one way to get it to work, but as for the idea itself -- I think I'd just know that the word is coming up because of the conversation as opposed to the actual word itself. Essentially memorizing the conversation instead of the vocab.
Although this is the same problem as with sentences, and one reason I go Vocab -> Definition/Sentence/Etc, but that's just personal choice, so whatever floats your boat

Yes, this will definitely work if you are interested in only one fact. But let's suppose we have a 2 min scene from a movie, out of which we make 5 cards just like you said. The SRS will schedule this 5 cards in random order. I, on the other hand, would like to look every time at the 2 min scene, with stops only at the cards that I have to review now. This will be just like looking at the movie, but with some stops because of SRS.