They had a post on their forums the other day abouot how newly submitted scripts would be evaluated by AI - I’m like, seriously? The same AI that sucks writing scripts is going to evaluate scripts now? I guess the scripts will have to suck to pass.
3 Likes
You and me both; And my scripts have always received good feedback here. Not worth it and the interactions are absolutely horrible on the PR side of things. Terrible company with a terrible format.
In my view scripting, as a service, should be organized for 3rd party hosts/producers. Then you wouldn’t have all your scripts for SwallowBay just up-end and disappear when the contract gets fucked. 
4 Likes
Haven’t you heard? Monthly subscribers are now known as abusers for thinking they can download up to their download limits. The preferred SLR customer apparently pays for the subscription then never uses it. Seems like the real response here would be an open source AI scripter and that would get around POVR, NA, and SLR’s ish here. Also, support CzechVR, BadoinkVR, and VRCosplayX for not being doodoo heads about scripts.
9 Likes
I had the same experience.
-
Extremely long wait times for them to “review” my scripts(21-30 days)
-
They were always nit picky with their complaints despite my scripts getting good reviews on here. They claim it’s because they hold scripters to an extremely high standard, but I am a script subscriber to SLR so I know they’re full of shit, the quality varies greatly from one scripter to another
-
Communication in general was slow, any email I sent them too a minimum of 48 hours to get a response, even simple yes/no questions that I sent within 5 mins of them emailing me.
Honestly it just felt to me like they didn’t actually WANT new scripters, but they had to keep up the facade for their subscribers that they were super duper working hard to hire human scripters to make high quality scripts.
7 Likes
Glad it’s not just me then…
4 Likes
Given he’s charging the same for AI scripts as human scripts and creating duplicate AI scripts for scenes already scripted, his goal here seems clear. Also, just watched some new AI scripts yesterday and while they have come far, they still aren’t up to the quality of what you guys make in any way. But the real cherry here is the way he cut the premium download quota for them from 10 to 3 probably because he sees anyone making use of their allocation instead of just giving him money as “abusers” and rants on and on about being the real victim.
10 seemed about right because they are really hit and miss even now, it’s just that they were mostly miss previously. And apparently he was OK with 10 when they were really bad, but as soon as they became OK, suddenly he put them on par with human scripts. And he justified 3 human scripts because he had to pay humans, so apparently now he’s paying AWS?
My question to you guys is did you sign away your rights to have your scripts used as training data for his model?
4 Likes
This is EXACTLY the impression I have had for some time. It is not very hard to read between the lines here.
3 Likes
keeping my fingers crossed someone comes up with an open, AI script generator. that will level the playing ground and hand scripters will get their due respect then. i don’t think it is far off and i would not mind contributing my scripts to train an open source AI script generator.
4 Likes
IMO, they’ve killed off their human scripting market; mostly PMVs and the like which I could not care less about. Only been premium for maybe six months (avoided that particular upgrade due to their businss practices and now regret it) Either way, when the first of the month came around, those 3 free human script tokens were immediately used, now we’re sitting at halfway through the month and I still have all three ready to go. And this has been the case for the past two or three months.
1 Like
I’ve answered this somewhere before, you can look it up, but I’m lazy - so I’ll just repeat a couple thoughts.
Currently any AI works only at the level of accuracy of 90-95% maximum, if you add to this level still correspondence of movements on the timestamps and lack of imagination in the AI I think SLR accuracy at 80%. The guys are great and have come this far in a little over a year, but to get the remaining 20% they need to spend 4 more of the same budget (and time) minimum, according to Paretto’s law.
Someone sold internally to investors on the idea - we’ll do it and do it better than other studios. SLR is not a research organization, I’m sure they didn’t set out to try AI in scripting - their goal was to do it. They don’t have the finances like Google or Meta to buy expensive equipment and hire engineers to try AI. I have no idea how much and what kind of equipment they have and how many engineers are working on this - for example they spent $1 million. It’s time to pay with interest, e.g. $1.3 million, but it’s up to the users to pay so SLR engineers can move on.
Will it work? I don’t know… Let’s keep watching.
2 Likes
He 100% could have taken an open federated approach to training the model across 1000s of subscriber machines in exchange for access to all the scripts that were generated with it. He seems to want to corner the market here, maybe sell to other VR sites, and stop relying on humans for scripting or even rating the scripts.
But the big mistake he’s made here is to set a benchmark for AI scripting with plentiful examples of the outputs. It will get duplicated and/or beaten, settle down with a bag of popcorn and the let the games begin.
3 Likes