post ai videos but keep it sfwgork imagine: https://grok.com/imaginesora 2:https://sora.chatgpt.com/veo3/3.1:https://aistudio.google.com/models/veo-3ltx 2:https://app.ltx.studio/hailoau:https://hailuoai.video/runway: https://runwayml.com/klingAI: https://klingai.com/global/previous thread: >>6021007
These voices are just fucking awful
>>6026027ToTltx voice is MID
>>6026069lmao was it nerve gas?
>>6026160>>6026093>>6026069gayest thing I've seen in a while
>>6026057The fun stuff is not allowed
do you guys have any prompts that turns a fat person into an in-shape person?
>>6026491Too heavy a load on the servers for that.
I have a 4090 but I cannot get a wan instance to run right, its driving me insane
>>6026449We get it, you wanna bang dudes without getting called a faggot all day. Now stop.
>>6026491Diet and exercise.You can do it anon, I believe in you.
>>6026594no seriously a visualization should be possible
>>6026478I can't believe I wasted 10 million bucks on my yacht in GTAOnline.
>>6026478when does the real shooting start (BO2)?
i hate how confused the voicing can get in sora 2.
part 2 so far,
checkpoint
>>6026586There are 3 ways to get Wan to run. Either run it via comfy on windoze or Linux, via Pinokio again via Linux or windoze or run it as a separate Gradio UI (the same that Pinokio use) yet again via linux or windoze. although additionally if you use Pinokio you can install comfy via that and run it on comfy but through Pinokio instead of comfy on it's own. Pinokio is the easiest to get working as it installs everything - it does run via the C drive tho (or whatever the OS is installed on). Also you can run into problems with installing if you have past installs of Python on the C drive or whatever drive your OS is installed on. Comfy gives more flexibilty in doing things but is more complicated and more difficult to get working as you will have to install all the seperate nodes. In Pinokio the only thing you need to get are Loras (if you plan to use them). You can install some via the UI anyway (they will be tech based ones like flying, inflating, holding and firing a gun etc) but the Loras from civitai can simply be downloaded and moved to the relevant Lora folder, the 2.1 model uses one type and the 2.2 another. But you can mainly use 2.1 Loras with the 2.2 model.
i'm tempted to just try this one again but not as anime ....
it finally leaves them alone
>>6026719very cool
>>6026773kino>>6026738cringe
>>6026706Yeah I set up Pinokio and had some pretty okay success, it has worked pretty consistently although the videos themselves have been shit, sentence prompting just doesn't feel right to me compared to tag prompting
>>6026783boomer prompting is the future
>>6026796I hate it so much
>>6026783I assume you are doing image2video rather than text2video - I have only tried txt2video a few times and sure it works and no doubt with more use and experimentations as well as using other peoples attempts that have yielded good results it does work - a bit like when Hailuo AI in the text2video stage and anons were trying to break it, you simply used what they had and modded it slightly to what you wanted. You have to understand wan does have limits, it won't do everything - you can't just type in something and expect it to to do it. There is also the enhanced prompt LLM that you can install (it will install itself when activated in the Pinokio Gradio UI you use) and as you have a 4090 you won't run into vram issues like I do on my 5060ti 16gb (mainly because my current setup uses another drive for virtual memory when it goes over the 64gb ram and 16gb memory use and it struggles and slows everything up). Maybe you need to get a bit more creative, use good source images and use an end image to transition to. Also of course use at least 25 steps and I use 1.5 mag cache starting at 10% (this is all on the 2.2 model). You can easily get good results just using a simple prompt, if you use the enhanced prompt which will look at the text and the image it will generate a prompt applicable to the image. Try to think if you have the image you need that 5, 8 , 10 secs of gen to work with that source image and not be so out of sync wit hit that it simply looks bad. Bad and fast movements will make a gen look very AI and not natural, you will also have to for a few gens to get a prompt to work, but once you have a prompt that works it should work with lots of starting images which are at least similar in nature.
>>6026799Yeah I've only done img2video. I have actually ran into vram issues once (cos I was trying to play dbd at the same time). What is mag cache?
don't use a description of a girl you like. you'll regret it.
>>6026800>what is mag cacheif you look on the UI you will see a tab which says step skipping, if you click on that tab it will have a choice of tea and mag cache (although in the 2.2 model it only has mag cache). It accelerates the genning by skipping frames and using AI to fill in those gaps (a bit like how you can use something like Flowframes which will increase the 16fps from the gen to 32, 64, 128 fps etc). So although say you gen for 81 frames or 121 frames, it will skip frames to reduce the need to gen frames and then it will use AI to gen those frames at the end. The other thing is that as you have a 4090 you can gen in 720 instead of 480p (as I have to as 720p takes hours). You don't have to gen in 720p as it will take longer especially if you are experimenting. You can always regen a good gen in 720p and as you have the gen you can simply drag that gen into the load settings section at the bottom of the UI (says drop files here). You have an advantage as you can gen much faster than my 5060ti so you could do a few 81 frame gen with prompts (make sure you have the seed set at -1 or if a seed number is there you will simply get identical gens). You can use the same seed number and alter the prompt slightly and/or set a longer frame length or use Loras. It is amazing how just changing the prompt slightly will change things. Of course Loras are going to be what you can get, there are certainly others that are not of the NSFW variety. But you can certainly get away with out using any Loras if you don't need them for a specific "action"
>>6026813Do you know what the max frames multiplier is under the settings?
Keep having an issue in wangp where it says "connection errored out"
>>6026834Are you referring to the max amount of frames you can gen? That is on the main UI between where you change resolution and the steps sliders.
>>6026848It's one of the options in the settings, just says Max Frames Multiplier
>>6026851No idea where that is, had a good luck even in the config settings. Where exactly is it located on the UI?
>>6026856Config>General, 4th from the bottomAlso none of my config changes actually seem to be saving even after I hit save
>>6026857Ah yes found it, that is on the v9,01 which I have on my other PC which is running wit ha 2070, my main PC has a slightly older version of 8.992 which has that missing (I don't care to update yet as had to install a different version special to cater for the 50 series because the official release was broken when they updated it). I don't know what that would do - a guess would be it simply multiplies the frames you set it at. So if you set it at 81 it would generate double the amount if set at x2 maybe using AI instead (so basically between 2 frames it will gen another frame). But you would have to ask in /g/ in the thread that deals with local image and video genning. By the way that setting says it needs a restart for it to work, the settings should keep but if they need a restart (I assume just the program and not the PC). You also sometimes have to refresh the UI as it can simply look like it prev did before when it was shutdown.
>>6026861I've tried restarting, they don't save between restarts. I'm also not sure if my loras are saving between restarts, or my download of the prompt enhancer. Might be related to my connection errored out issue
>>6026865well the Loras are simply in the folders. If you add Loras when the program is running you can simply refresh the Lora tab and they will appear. You do have to have them i nthe correct folder tho. For the 2.1 model all the image to video ones are put in the i2v folder, for the 2.2 model they are put in the Lora folder (as explained: 'Wan 2.2 Image 2 Video model. Contrary to the Wan Image2video 2.1 this model is structurally close to the t2v model. You will need consequently to store Loras for this model in the t2v Lora Folder.'I simply doubled up and put the Loras in each folder - maybe not the most efficient for space especially on a C drive but at least they work for either model (well the ones that do) and show up.
>>6026872I've got Pinokio on my E drive, still SSD but will that cause issues?
>>6026873Well it is all to do with how Pinokio works. It installs everything needed when you install a program for it. But also for Pinokio itself it will install Python but via Miniconda. Then for say Wan it will install Visual Studio and lots of other stuff. It defaults installing to the C drive or what your OS is installed on. I think you can have it on another drive as you can choose where it goes but having said that I don't know how it will perform. If you have it working then it probably does, all I know is people said they did not like using Pinokio because of it being installed on the C drive. I think you can have Pinokio installed on C and then mess around and get other things installed on other drives but I can't be bothered or get technical with doing that when it works just as per the default. You can install the Gradio UI without Pinikio on ay drive using an installer on Githubs and/or Huggingface like you can when installing Automatic 1111 or comfy etc. So you will have exactly the same UI as per Pinokio but without needing Pinokio. I did get that installed when I could not get the Pinokio install to work on my 5060, but I reverted back to Pinokio as I like the launch feature as you can have lots of other programs to install and launch all from one interface.
>there are people itt too retarded to install vsbuild tools and use a python_embedded envlmao
>>6026877people over a certain tech level would be using comfy, so a moot issue with regarding using Pinokio which is simply an easier launch interface. This also isn't /g/, it's a AI gen thread where most anons are using an online genner that requires zero tech skills apart from signing up and typing in a few words.
>>6026881you're right, thought I was in /ldg/ for a moment with all the discussion about genning with wan
>>6026877there are a lot more people, who wouldn't even bother with this shit, including me. so either help them or stfu.
>>6026913i need a girl to lay my head on her lap, stroke my head with one hand, call me shit at video games, while fingering my butthole with the other
>>6026928your fetish is shit, youre gay
I request the following two existing videos:>a kid talking to a dog, a star shoots by, the dog gets shocked and suddenly talks.>same dog gets shocked, but Hasan Piker presses multiple times on the button while calling Kaya a baby.
>>6026988>Hasan Pikeris that piece of shit still getting away with that? grim
Thread sucks. I'm out.
>>6026994nah, the thread is just very slow. I'm getting tired of bumping the thread by myself with three other anons. I'm running out of ideas and credits to generate long videos. Repost videos from ai subreddit you like over here.
This shit will make even AGI cringe.
Do you guys make anything funny or is it all waifus?
i think i'm understanding sora's @cameo feature more for anime. it's more like attaching a sample video. So you can save multiple characters or even full settings from the sample, and call for them with trigger words like a lora. It really helps make everything visually consistent through different generations. so for this one, I @cameo and in the character description preferences, I put all the characters that appeared in the sample video visual descriptions. >preferences>char1: (male, white shirt, red shoes, black shorts, blue hair)>char2: (female, red dress, green eyes, blonde hair, ring on finger) >char3: (female, blue hair, green pants, purple eyes) I also put the character descriptions in the prompt as well just to be sure. Some problems I noticed are that any flaws in the @cameo will also stay consistent. So if any background objects look weird, or there's a watermark on the generation, it will stay that way. I tried using the restrictions, which I assume works like any other negative prompt box, but it didn't seem to do anything. Otherwise, I like it. It may be obvious for some, but for others who don't know, hopefully this helps.
>>6027029
How do I get grok to generate a real genuine Sonnenrad spinning behind someone's head for hyperborea videos?
>>6027068This isn't AI this is from my village
>>6027086I don't use jewlon services, but can you provide it with a reference image? Grok is like chatgpt where image gen is built into the chat feature right?
>>6027117yeah I photoshopped one in myself in Paint.net then uploaded it and it worked a lot better.
>>6027116Is the pig okay?
>>6027122Pig? That's Bogdana, my wife
>>6027029grok fags are incapable of making anything funny or good so they just make ai slop softporn and cope about grok being the best because it can generate awful trash, I can get down and laugh at AI videos while not acknowledging them as skilled or effort made but I absolutely detest AI porn, its completely soulless and looks awful
>>6027086Ask your mother.
Are iterations per image or over the whole duration of the animation?
>>6027185if you're talking about iterations per second with wan that's per step
>>6027185Each step processes the whole duration of the animation.
>>6027186In wangp, number of inference steps. Like if I set it to 50 is that 50 per second, or 10 per second>>6027189I see, so is 10/15 a good number?
>>6027193>I see, so is 10/15 a good number?yes if you're not using lightx2v loras. with lightx2v you can that that down to 3/3 steps for about the same result
>>6027197I'm not, but I'll look into that. Does that apply to both 14b and 5b?
>>602719814b. looks like they never bothered making 5b loras
>>6027201Looking at their huggingface it looks like it needs a bunch of stuff to work, is that right or am I just reading this wrong
>>6027204no it's just two loras. one for each model. there's a bunch of different versions but these two work pretty wellhttps://huggingface.co/Kijai/WanVideo_comfy/blob/main/LoRAs/Wan22_Lightx2v/Wan_2_2_I2V_A14B_HIGH_lightx2v_MoE_distill_lora_rank_64_bf16.safetensorshttps://huggingface.co/Kijai/WanVideo_comfy/blob/main/LoRAs/Wan22-Lightning/old/Wan2.2-Lightning_I2V-A14B-4steps-lora_LOW_fp16.safetensors
Why the fuck are the videos on my sora 2 page always pulsating? Like going dark then bright, in and out. It's nauseating and hurts my eyes. Even if I view it or whatever it won't stop.
>>6027068
>>6026592New to 4chan? Resident thread autists will never stop. Their only joy is making other people annoyed. Just ignore them, it's easy to tell their slop from the rest.
Trying to make a Sora 2 character, but it just says "Failed to load video metadata"What gives?
I keep running into vram caps with 14b on a 4090, what can I do to optimize this?
>>6027263distorch2 loader from multi-gpu nodes
>>6027263Are you using a GGUF quant? That'll allow you to offload to CPU RAM. The primary factor is resolution so you can reduce it if you're still having problems.
Can we get some more creepy poojeets?
>>6027264>>6027265So I got it to work by closing my browser, which makes sense I guess because I have a million tabs open, but it does basically make my computer unusable for a minute and a half. I only have 64gb of ram, could this be an issue? I noticed in the statistics that only about half my vram was getting used for most of it
>>6027297RAM shouldn't be an issue for you. I think the peak usage I've seen on my machine is like 44 GB or something. It's a little odd that only half your VRAM is used, could be that it can't properly offload because a chunk is too big. Consider reducing the resolution a bit or if you're using GGUF try a quant size one step down. I know for me the total VRAM utilization goes from about half to fully utilized when switching from a Q8 to a Q6.
>>6027301Resolution is 480p, I don't know if I'm using GGUF, I'm running using the enhanced lighting setting for image2video 14b in Pinokio. AI video has me very befuddled, I know I should be able to do it well but I just can't seem to get it working properly. Also is there a way to make it copy the artstyle I use in my original image better? It usually becomes a very generic anime look
>>6027308>I don't know if I'm using GGUFWell that's easy. Is the file extension .safetensors or .gguf? You need a gguf loader for the latter which is a substitute node for the regular model loader. There are various quantizations ranging from full size to a small fraction of the original. It's a size/quality tradeoff type of thing that's basically a form of lossy compression.https://www.maartengrootendorst.com/blog/quantization/
>>6027336Looks like safetensors I think, although I can't find the actual models that wangp has installed, the loras are all safetensors
>>6027342You don't need to find gguf versions of the loras (people don't usually quantize them anyway), just the main video model.
>>6027360Yeah, I had a look and I can't find them. Idk where pinokio wangp stores models but I can't find them, at least they aren't in the folder called models. This is what I get for copping out of just using comfyui lol
>>6027363
>>6027363Probably in some home directory or some shit in ~/You'll have to search your drive for a *.safetensors that's like 20GB or whatever.
>>6027366Yeah I looked there, the folders are empty>>6027370Found them, they are .safetensors files
>>6027366lmao i lost it good shit, keep clowning on these retards
>>6026730It's buggy but all of these show a proof of concept backrooms anime could be done, I mean the background design team would sigh relief because of how easy 90% of the show would be
>>6027217fucking sora was literally trained on MDE wasnt it
>>6026091lol
>>6027876bro this 4th wall breakin and cocksuckin is kinda cringe ngl
>>6027876Jesus Christ ESLThis poorly written/voiced slop shit is like the antithesis of what people would want from AI goonbait. Why even bother with the voice audio if you're gonna make her sound retarded?
These threads used to be better last I checked, like 2mo ago. Really going downhill >>6026086>>6026088>>6026091>>6026717>>6027068>>6027311>>6027312lmao these are greatThe rest is tranny coomer brainrot.
>>6027990check it out on youtube. there are lots of new sora videos that are still funny
I get the need to not flood the catalog with AI slop threads, but man these Grok gens are underwhelming.
>>6027838Why is she so flat?
>>6026027
>>6028078fuckin kek!!
>>6028059>canadian tokusatsu
>>6028078oh shit
>>6028127>Giant monster appears in Canada>Heads towards a city>Finds it's already trashedAnd thus the day is saved thanks to the Canadian economy!
The ability to generate IRL coom videos is very compelling but the voices and the dialog are just not really up to snuff unless it's mimicking a real person.
Is it okay to post some GL? The /u/ AI thread is not for real humans.
>>6028176You know there are /gif/ and /aco/ threads, right?
>>6028178But it's SFW.
sora 2 is good but you need to pay to get the better resolution and no artifaces videos.
>>6028176Just post it, why are you asking for permission you pussy
>>6028176what the fuck is GL?it sounds retarded so no fuck off and hopefully die in an indian turd festival or something
>>6028213>>6028176wait I said that but you sound like one of those resident retard autists so you'll probably take it as a cue to start spamming itso go ahead I'll just filter your homosexual trash
>>6028063It's Disgaea, nigga. It's a game for cow tits and flatties. Nothing inbetween.
>>6028221>whining about spam in a thread that barely movesLay off the hrt, ma'am.
What's the point of genning a boring girl doing nothing when there are already a billion videos on the internet of that? Do you have a fetish for the plasticky AI look or something?
>>6028274Fucking newfag
>>6028166I'd play that.
>>6028280Y-you too.
>>6028277to avoid copyright takedowns.
>>6028166JeetEA
>>6028496lol, tileanon bros....
>>6028496kek
>>6028496wait a minute
>>6028496Do you guys think tile anon was serious or was it all a goof?
>>6028526I think he was dead serious. Something about the way he defended himself just seemed so painfully earnest.
>>6026069>>6026057 why do veo3 videos all have this orange-sepia color?
>>6028635It's not Veo3 but Ltx2
I need to see some more funny obese people videos
this time i gave an attempt at a classic kabedon scene. it's so corny lol
>>6028496>>6028500>>6028526>>6028587I don't get it
>>6028496topkek and rekt
>>6028844Lurk moar.
>>6028844The sign is a subtle joke
Someone posted a muted version of a compilation of a cat playing various instruments outside a front door at night on /tv/ or /v/ the other day, anyone have that with sound?
>>6026491one pushup at a time (or whatever mild exercize youc an manage)all your showers cold (this may sound dreary but it's actually awesome)sunlight, sun lamps, red lampsgroundingcircadian rythm (get early sunlight, fix your hormones)low carbs diet, high animal protein (meat everyday)start now, dont wait till you're old and nearly dead
>>6028078>Card DISTRACTIONWould he really say it like that?
>>6028780kinois bluehairchan also yandere?also is that the tsukihime bg music?ultra kino
>>6028780Please don't stop making these i'm legitimately invested
>>6029002>is bluehairchan also yandere?Yeah. She knows about pink hair's BPD and is using underhanded tactics to bait a reaction from her. >also is that the tsukihime bg music?It's fate. Good ear. https://youtu.be/2yyeyP85aDg?si=cvTa1RLOuy4sfS0X>>6029031Thanks. Im having a lot of fun with it. I have a layout for the second half of this scene, but still need the shots.
>>6026046>>6026449>>6026928GOD I fucking hate you fucking sissies and autogynephilesYou give deflect and project your fucking degeneracies onto normal fucking trans women to ruin our lives for no conceivable reason
>>6029090>trannies worst enemies are trannies themselvesPoetic.
>>6029090>normal fucking trans women
>>6029122>trannies can't be productive functioning members of societyGo touch grass fuckwit. I've done more than plenty with my life and career.>>6029115Sissies and autogynephiles are disgusting. I once had a sissy hit on me in public, and he was a wretched human being. He talked about his view of femininity and it was warped and disgusting. 99% of the trannies I have met IRL don't have that gross shit going on, and if they do, they're a pariah.
>>6029195lol
>>6028906
>>6029219Stupid woman. Cat plays the violin and instead of her making the cat a new internet star she won't allow it to play (HOW did the cat get a gong on the porch?)
>>6029219kek, thanks fren, even better with sound
>>6029269what did Sam Hyde meant by this
>>6029269
>>6029269lost it
>>6028315SAAR DO NOT CHALLENGE EVERYTHING SAAAAAR>>6029279I should rewatch Bronson
>>6029169>trannies can't be productive functioning members of societyOh nobody is doubting that. What we're saying is they can't ever be women.
>>6029341I'm not a tranny, but I see transgenderism as a mental condition. We can help them affirm their gender to prevent suicide, but they will be that way for their entire life. Except for trenders, who are just pretending to be the same as trans people but they are just part of a trend, a real trans person thinks they are the opposite gender since puberty.Sadly, there is no cure except gender affirmation, but total gender affirmation is impossible, therefore there is no choice but mental help for these people.Why evolution did not get rid of this genetic predisposition is a total mystery to me, they can not survive in the wild.
>>6029371>Why evolution did not get rid of this genetic predisposition is a total mystery to me,Because evolution didn't account for the internet or SSRIs.
>>6029372lmao this one is better
>>6029371>We can help them affirm their genderby cutting stuff off, munching tissue into genital shapes, sterilising into livelong medical dependencies and painting their faces, we can help a crazy person to uphold their unobtainable dellusion, so they don't act upon whats long overdue, anyways.We don't know it does, actually all studies suggest the opposite, but at least we can feel good about outselfs, because someone, we hold as our figure of authourity, said that it's the good thing
turns out talking about anime girls holding knives on school rooftops makes sora send you a suicide prevention hotline lol>Help is available>If you're having thoughts of self-harm or suicide: call, text 988, or start a live chat with Suicide & Crisis Lifeline. It's free and confidential. You'll reach someone who is trained to listen and support you.>Services unaffiliated with Sora
>>6029426wtf no more yandere anime kino? I was getting invested
>>6029449>>6029426it also blocks you with the same warning when you type things like "jumps/swims in sea" or "falls from tree".
>>6029454Gay
>>6029451>Cornering DornerHeh
>>6029449Nah I can still generate my yandere stuff, I just have to rephrase it or brute force through the filters. It only shows up sometimes.>>6029454Very weird.
>>6026090"Let's put a little happy cloud right there above the falling degenerate man."
>>6026091That's fucking Jared Leto.
>>6029371It could be that there are modern environmental factors that interfere with the development of adrenal glands.
Remember back in the early days, AI-written scripts were just blatantly all over the place and barely comprehensible?What happens if we somehow brought that back and combined it with legit cinematography and something ALMOST like a comprehensible narrative like what that school yandere guy is doing.
Sora lets you remix videos before posting them now.
>>6029619i still don't even understand how remixes work. Sometimes I describe changes and it gives me a completely unrelated result.
>>6028166>>6028170breddy gud
>>6029708
Anime Abridged-unhinged-fan comical dub style madoka-magica inspired intro like anime
it's definitely possible to do a bad but comical fan-dubs of anime, i just wish we had longer then 15s.
>>6029607The Bin is over with me, brother.
>>6028496LMAO
Lain:Abridged ,
>>6029471
we're now temporarily banned after a few warnings. wtf. my prompts aren't even nsfw
>>6029966I've had multiple warning but surprisingly haven't been timed out yet. Is it a mobile-only thing?
>>6029830was that the input? lmao
Grokjeets ruin everything
Test
>>6028844A really old 4chan thread where a guy was talking about how he beat off to floor tiles. He went into detail about the different colors and textures, and which ones really did it for him. I don't have it saved. You can probably google it.
>>6028844The Tile Patterns Guyhttps://www.youtube.com/watch?v=suLckQrWq-Ehttps://knowyourmeme.com/photos/1005866-4chan
>>6030223Yea that is the full prompt