at a crossroads about how I should tackle subs2srs.

Right now, my front card is the audio/text.
Problem is, I usually remember every scene from shows I watch. Thus, when I hear the audio, I just remember the sentence from the show's scene. I can remember the sentence due to context, not vocab. However, if I had to recall individual vocab from a sentence, I don't think i'll be able to do it..

How are you guys using subs2srs effectively?
just memorize the sentence or have to be able to recall individual words/kanji?

That's a feature, not a bug.
You're doing recognition, right? Recalling words from context is great, even if you're thrown up a totally different context. If you run into that word somewhere else there's a pretty good chance you'll recall it, quite probably in the voice of the character you were subs2srsing.
Though of course, if you're expecting to recall it in order to say it - that's unlikely to happen. But I don't think you should expect SRS cards to do that for you anyway.

I think memorising the sentence is a poor idea, if you mean that literally. If you understand the sentence when it comes up, that's good enough. I used to listen+read, then repeat - a pass was if I both understood the card and could say it. I still think it was a good way to go.

I second what Aikynaro said. The purpose of subs2srs is to get better at understanding native speech and colloquialisms - not learning vocabulary. Of course you will learn vocabulary along the way, but I don't believe subs2srs is the most efficient method for vocabulary acquisition.

I rather disagree with that is being said here, mostly through my own failed experiences with trying to learn through sentences and context like this.
I usually just learn the sentence and don't have a clue about the individual words making it up. If I see a word used in a different context, I don't recognize it, or I might recognize it but not be able to remember anything about it.

So I make simple collocation cards such as "Object Particle Verb". Or if I feel that a word can really just stand on its own, then just a simple vocabulary card. The idea is that I put as little context as possible on the card while still having enough data that the target word or phrase is not ambiguous in any way. This ensures that I learn what I am setting out to learn, rather than just recognizing some specific instance of it.

This is not to say that I think sentence cards are useless. They might be a good accompaniment to my collocation and vocabulary cards. But sentence cards on their own just don't work for me.

Xarxrax, I suggest you re-read the OP and responses. I'm in agreement with what you've said, but op is asking specifically about subs2srs cards.

I don't think its really any different. I've done subs2srs cards and they are basically the same as sentence cards + audio. What's the difference?

There's no difference. It's just that you didn't answer the OP's question and said that you disagree with something that I think we actually agree on. If it sounds like I'm spoiling for a fight, it's not my intention. It just sounded like you weren't responding to what was written. It happens sometimes. No worries. I only wrote so OP didn't think he should turn his subs2srs cards into vocab cards or something.