Reading your comment promoted me to go have a look at the request page and I do see the cause for concern as for example I see a new video that was only dropped two days ago(Maid Surprise - SLR Originals - VR Porn Video | SexLikeReal) has like 78 votes while older videos will indeed suffer from things like
a) people not knowing about this new feature
b)people who loved the older videos no longer being subbed to SLR to vote on them or
c) people being overall lazy and not wanted to go searching for every video they like and hitting request
A couple ways to help would be
a) get this new feature in peoples face and use it in advertisement( but i think they want to get their AI scripts perfect before they advertise it.)
b) Maybe have a new tier of subscription( like Tier 1 - Videos + Crafted Scripts, Tier 2 - Videos + AI Scripts, Tier 3 - Just Videos)
c) make requesting videos in masse easier by having like a checkbox and a list format so I can go down the list check the boxes and when I’m ready I can hit request all.
Anyways sorry didn’t mean for it to be that long but I just had 1 idea after another and I’ve been subbed to SLR for a long time(bout a year) and 1st time paying for porn so I had a lot of thoughts
The goal is to have the AI create scripts without human intervention. The AI scripts we currently post on SLR are 100% unaltered. (NOT adjusted/tuned manually.)
Both Realcumber and myself review the AI scripts. Then our AI team makes changes based on our feedback. But we love to hear from the community as well! The scripts are after all meant for the users.
I also want to clarify, the current AI scripts are not final versions! They will get replaced as the AI gets better!
Thanks for the feedback - it’s nice to be part of the journey!
Review time - I did get curious and purchased the Emma Hix AI script. All in all, it is better than other AI scripts I’ve come across (Adult time comes to mind), but still has a LOOOONNNNG way to go to get even close to a halfway decent scripter. Lots of missing nuance and stroke lengths don’t feel as good as a manual script. There are aspects of manual scripting - like exaggerating stroke length depending on the intensity of the performers or during “act 3” of a scene - that AI has some work to catch up on.
To the community - I’d recommend against purchasing these scripts “a la carte” for now. If there is a scene that you REALLY want, or the hope/expectation that the script is replaced in the future… mayyyybe pick up a script. But for those of us with a script-membership from SLR - by all means, give the AI scripts a try. It’s pretty neat to see the evolution of our hobby.
Here’s an example review of an unpublished AI script due to safety concerns.
01:24:42 to 01:29:42 Bunch of potential unsafe parts.
The AI didn’t seem to know what to do with the toy part at the beginning. [REDACTED]
Inconsistent stroke depths, but overall better than prior batch when strokes were too shallow.
Blowjob tracking shows big improvement compared to previous script batches. The inconsistency mentioned above, seems to be closer related to [REDACTED]. AI is taking distance between [REDACTED] and [REDACTED] into account, [REDACTED]
The timing of the dots (strokes) is a lot better compared to previous batches, but is in some places way off.
Implementation of making strokes less shallow, seems to have gotten unwanted results for [REDACTED], which were pretty good in previous batches.
Examples:
00:20:08 - There is a delta of 44 stroke depth between two strokes that should have a delta in the range of 0 to 5.
- Stroke 1: from 98 down to 48
- Stroke 2: From 99 down to 4
By Delta I am referring to the difference between 48 and 4, making the second stroke almost twice that of the first stroke.
Visually these are pretty much the same.
00:20:35 - Good blowjob tracking, followed by two well executed swaps between lips, to hand, back to lips.
00:20:56 - I’ve seen by now good speed changes. Strange it is not scripted this way on this downward stroke. Stroke starts at 99 downward. When depth of 63 is reached, it should still be closer to 85, then speed up.
00:21:03 - Another example of good switch between hand and mouth tracking.
00:21:23 - Inconsistent stroke depth, but this time in the upward direction.
00:21:28 - Stroke too soon (noticeable in usage)
00:21:35 - Here the AI did a much better job going from slow to speeding up. (comparing with feedback timestamp 00:20:56)
Contrary to the feedback on script for scene ID [REDACTED], Here the fast large strokes are in line with expectations and visuals. However, [REDACTED], while the strokes are much more similar to each other.
00:26:00 - Another example of how good the tracking of head is, similar to blowjob tracking.
00:32:29 - Strokes a bit to large, making some of them too fast. But here they are not sudden and unexpected.
00:51:27 - This one is borderline. It’s unexpected and not in line with the visuals, but I don’t think will cause any issues.
01:02:52 - [REDACTED].
01:24:42 to 01:29:42 Bunch of potential unsafe parts.
[REDACTED] was very good in previous batches, but it’s nowhere near QA passable in this instance. Not just for the potential unsafe parts, it’s the whole segment.
01:31:10 - Potentially unsafe. not in sync with video, and unexpected.
01:32:31 to end - [REDACTED], Also all over the place like [REDACTED], but better.
AMAZING. Ya’ll are definitely doing your homework - Love the detail of feedback and the desire to keep improving your model when other sites are happy with putting out trash. I’m rooting for ya - the more quality scripts the better!
Hi people! It has been some time, but we took it to improve and today I would like o share with you 10 more AI scripts. Check them out and let us know what you think!
@Kieshi May I suggest not tasking out highly requested script scenes to the AI generator and instead using mid to lower-requested scenes? I think everyone would agree that highly favored scenes should have higher-quality scripts.
Haven’t tested them out yet, but this looks like a good mix.
There are newer and older scenes.
One-on-one and group scenes.
Some ethnic/exotic thrown in there.
As far as I can see you, you avoided those vids with extremely high request numbers like 90+. But you chose those a lot of people (70-80ish) wanted. Seems like a good approach.
I am a total Scarlett Alexis stan so it feels sacrilegious to complain about more scripts for her, but I think a better tactic would be to not duplicate actresses in these block script releases.
Thanks for all your work, I’m looking forward to the AI up-n-down with cheerleader Scarlett.
Nice, it’s definitely better than nothing! AI scripts are a lot better now, well done
Have you considered making AI scripts free to download? I would enjoy being able to download and edit them myself, fix a stroke here blah blah blah… I tried making scripts and I hate making scripts, I always wished I could get a solid AI one and adjust as needed.
I totally understand why not, you’re a company and need to turn a profit. Maybe for different tiers of subscribers?
First impression of the new batch is that they’re better in terms of accuracy with straight-up riding scenes. But that fine details were lost pretty consistently in BJ/HJ scenes (particularly whenever both were at play) and grinding was often a blank zone.
I tried about 3 of the new batch. IMHO they are worse than the last batch.
Unnecessary movements when no action is going on.
Missed strokes.
And this weird pattern where every second or third stroke was very shallow. Like Scarlett was in cowgirl (knees down) and just thrusting back in what I felt was a fairly regular pattern and depth. But for some reason the AI read this as deep-deep-shallow-deep-shallow-deep-deep-shallow. It was very off-putting.
My mission wasn’t a comprehensive review, so I may be mistaken. Oiled Massage from Scarlett Alexis around 36:30.