A thread dedicated to the discussion of AI Vtuber Chatbots./wAIfu/ Status: Space exploration!>Thread templatehttps://rentry.org/waifuvttemplate>How to anonymize your logs so you can post them without the crushing shameInstall thishttps://github.com/TheZennou/STExtension-SnapshotThen after you've wiped off your hands, take a look at the text box where you type stuff. Click the second button from the left side, then select snapshot, then select the anonymization options you want.https://files.catbox.moe/yoaofn.png>How to spice up your RPing a bithttps://github.com/Rurijian/Deep-Swipehttps://github.com/artisticMink/openrouter-roulette-for-sillytavern>General AI related informationhttps://rentry.org/waifuvthttps://rentry.org/waifufrankenstein>How to use Gemini with SillyTavernhttps://aistudio.google.com/prompts/new_chatSign in, then click the blue "get api key"Put it in silly tavern and voilaCourtesy of ERBird, Nerissa's most devoted bird and eternal player of GFL2.You want to leave the proxy stuff blank since you aren't using one when doing this.https://www.reddit.com/r/SillyTavernAI/comments/1ksvcdl/comment/mtoqx02>Other optionsMiku.gghttps://venus.chub.ai/Openrouter wants a one-time payment (think of it as a deposit) of $10 and you can get 1,000 messages per day. As long as you stick to free models you only need to put that much money into your account once.>A primer on getting voice working in Sillytavern (there are other options, just play around).[YouTube] Chat with Chatterbox in SillyTavern! https://github.com/devnen/Chatterbox-TTS-Server>Tavern:https://rentry.org/Tavern4Retardshttps://github.com/SillyLossy/TavernAI>Agnai:https://agnai.chat/>Pygmalionhttps://pygmalion.chat>Local Guideshttps://apxml.com/tools/vram-calculator[Koboldcpp]https://rentry.org/llama_v2_sillytavernWho we are?https://rentry.co/wAIfuTravelkitWhere/How to talk to chatbots?https://rentry.co/wAIfuTravelkitTutorial & guides?https://rentry.co/wAIfuTravelkitWhere to find cards?https://rentry.co/wAIfuTravelkitOther infohttps://rentry.co/wAIfuTravelkitYou can find already existing bots and tavern cards in the links below:>Bot lists and Tavern Cards:[/wAIfu/ Bot List]https://rentry.org/wAIfu_Bot_List_Final[4chan Bot list]https://rentry.org/meta_bot_list[/wAIfu/ Tavern Card Archive]https://mega.nz/folder/cLkFBAqB#uPCwSIuIVECSogtW8acoaw>Card Editiors/A way to easily port CAI bots to Tarvern Cards[Easily Port CAI bots to Tavern Cards]https://rentry.org/Easily_Port_CAI_Bots_to_tavern_cards[Tavern Card Editor & all-in-one tool]https://character-tools.srjuggernaut.dev/>Some other things that might be of use:[/wAIfu/ caps archive]https://mega.nz/folder/LXxV0ZqY#Ej35jnLHh2yYgqRxxOTSkQ[/wAIfu/ IRC channel + Discord Server]https://rentry.org/wAIRCfuscordMatrix>Lorebook management stuff[Worldinfo drawer]https://github.com/lazuli-s/SillyTavern-WorldInfoDrawer?tab=readme-ov-file[Standalone editor]https://github.com/ActualBroeckchen/SLEdPrevious thread: >>110589643
Anchor post - reply with any requests for bots, with your own creations, or with your thoughts on the enshittification of life.You can find already existing bots and tavern cards in the links below:>Bot lists and Tavern Cards:[/wAIfu/ Bot List]https://rentry.org/wAIfu_Bot_List_Final[4chan Bot list]https://rentry.org/meta_bot_list[/wAIfu/ Tavern Card Archive]https://mega.nz/folder/cLkFBAqB#uPCwSIuIVECSogtW8acoaw>Card Editiors/A way to easily port CAI bots to Tarvern Cards[Easily Port CAI bots to Tavern Cards]https://rentry.org/Easily_Port_CAI_Bots_to_tavern_cards[Tavern Card Editor & all-in-one tool]https://character-tools.srjuggernaut.dev/
Also i just not realized i fucked up the subject and put it into the name field.
Word Cloud for the previous thread
So they're playing Hytale now.
wud u dood a moom?
>10
https://www.youtube.com/watch?v=-bL9Smr4JXk
If you guys haven't and get the chance to, try roleplaying with Gemma 4, either 26B-A4B or 31B. In my opinion, it can trade blows with Kimi and GLM. Haven't seen a model this size absolutely mog models multiple times larger at a task for a while. It's quite good with translation too. Spent the past week losing my mind and having fun with it. Still waiting on Deepseek v4 but rumors are getting a whole lot stronger especially after the website was overhauled.>>110797264I gave it the old college try to write her a card but I haven't done it before trying to do this with some transcripts of the video and her interaction with chat. It kinda is a bit too tryhard and moody to be Bibi but that could just be because of Gemma 4 being very railroaded but doubtful since it was great with some of the other cards I had. Hopefully someone can improve on it.https://litter.catbox.moe/oy192a44tx3l9s6z.json
>>110959412>either 26B-A4B or 31Bare there any significant differences between the two? which one works better for cooming? can you also send the preset you're using for Gemma?
>>110959970ntaA4B is a MoE model while 31B is a dense modeltldr:>A4B is a bunch of small models shoved into a bigger model and does magic codeywodey to pick which smaller model to use on runtime>32B is a fuckhuge model like your used to usingshoutout to Pygmalion burning so we may live.
>>11095997031B is technically better, but I use 26B because MOE speed..Even if you offload with a small GPU, as long as you have enough RAM, you can offload it at Q4_K_M at 14 GB total where a 32GB RAM 8GB GPU typical gaming system can run it really quickly. I'm using the following settings even though you can find a ton flying around.>Temperature 0.75>Top-k 64>Top-p 0.95>Min-p 0.05>Repeat-penalty 1.0>Presence-penalty 0.5>>110960148This wasn't Pygmalion although yeah, they deserved it. But most of the goodness is most likely the Character AI deal hollowing out that company which made Gemma 4 so good, if you recall:>https://blog.character.ai/our-next-phase-of-growth/Whatever secret sauce and knowledge from there was too late to put into Gemma 3 but boy oh boy is it undeniably here in Gemma 4. Only issue is that it's most likely distilled so it's a bit overcooked on logit possibilities and a bit railroaded in scenarios so if you go through the same basic options in an RP, it will largely play out the same way. There are smaller models but they are too small to really be useful outside of general contexts on a cell phone offline where it is really good.
>>110960465im surprised your not fucking around and training E4B locally since you (apparently according to a reddit shill post for something called 'unsloth') only need 10GB vram for thatthats kinda the only appeal for me over these releases but i really really do not want to curtail my own distilled deepseek dataset for that right now
>>110960148>>110960465so it's basically down to personal preference? i use OpenRouter so all this talk of GPUs and RAM and offloading doesn't matter to me
>>110960762E4B isn't enough to do anything useful for my purposes and I don't use Unsloth since I prefer axolotl but it takes time to get support in. I don't have the hardware to run training for the big boy models nor the money to rent it even if it was there at the moment. Also technically tied up with several things I have to get back to so playing around with Gemma will stop after today.>>110960929One is clearly better but it will cost you and the speed difference I mentioned is real.
>>110959412I’ve seen a lot of promising shit done by the Japanese with Gemma and voice Gen to chat with chuuba style waifus.I didn’t save any links because I want it to be more developed before I try it so go find it yourselves if you’re curious ;).
>>110964939>I didn’t save any links because I want it to be more developed before I try it so go find it yourselves if you’re curious ;).you should go check your mailbox anon.you should go check it right now.you should go check it before the mailman comes around tomorrow.
>set up Gemma 4 (free) on ST>429 Too Many RequestsGod doesn't want me to coom right now, i guess
>>110965000
when you chat with coom cards, do you base your dude's dick size on yours or do you change it?
>sex god panko
good night, /wAIfu/please don't put a Doom nukage barrel next to me and my bed before shooting it while i sleep
>>110968401Depends on if I’m playing a character that’s not me
>>110960762Distill what who what where what share the bounty.>>110960148What happened to them?
>>110976664>Distill whatcreate my own dataset to fine tune with by picking and choosing high quality replies>wha happenno idea but im assuming normal anon falling out bullshit
>>110976664The lead disappeared before Llama 1 leaked and the remaining devs all disappeared slowly and the ones who remained went full grifter mode with a website and commerical operations. It's small peanuts though and it's deserved for them remaining in obscurity although it is true too ragtag anons can't compete with Character.AI with former Google researchers. The main thing that pisses me off is that the public dataset, PIPPA, is so tiny vs the submissions they got and that they didn't keep trying to do local AI and gave up the dream of local models end to end. Well, no use crying over spilled milk, I'll take what we have now with open models even if we can't replicate it because we need millions of GPU time to reproduce it alongside no public data.
Heard about Gemma4. Don't know shit about LLMs, but I've been wanting to "train" a model on my favourite smut novels for a long time now and it seems Gemma4 makes that possible.Can you guys just confirm that the basics are sound? I've got an i7-12700K, 64 GB RAM, RTX 3080 with 10 GB VRAM.Which of the Gemma4 models is best suited for this type of training that I can run/train with my hardware? Quality over speed for the output.
>>110981150How hard is that to do and how come no one is sharing their distilled something something on GitHub if it’s visible for us plebs to do?
>>110983890>takes more (passive) time and money than anything else since you need to let a gpu go brrrrr 24/7>need a reasonably new gpu / googa notebook or else it will take a lot of time to actually finish>because most people are using shit prompts and getting shit responses, so training on shit replies is just going to make things worse for the average ahh ahh mistress>the average person benefits more from paying to use deepseek/[big corpo model] and leveraging its innate ability to unfuck a prompt though sheer intelligence more than they would using a smaller local model>you need like 2 gigs or whatever of text so its kind of a lot of data to horde just to fine tune>no guarantee the fine tune will be better>"fuck you build your own">you still need a reasonably new gpu to get good reply speed>finetunes are not magic improvementspick between 1 and all but its not that hard in reality, just annoying to the point where the effort is not really with the juice.realistically the use of a fine tune is to create a LoRA for an updated dataset of something that was 'created' after its cutoff datelike how deepseek 3.2 doesnt know who smugalana or veibei are, a fine tune is worth more to 'add' native info about something for this use case than to crank out "better" replies since raw info about people is easier to source than hundreds if not thousands of (high quality) replies>see pic rel, ignore the 'edit' its the only way i can regen a reply in bravean alternate use case would be to have a model NEVER act as {{user}} in the reply without having to hammer it into them, but thats a lot of effort/time/cost when>do not act as {{user}} in your replyis "free" to do and gets you 85%+ the way therethe other realistic use case you might want is to uncensor a model but again, you would need a uncensored dataset to uncensor the repliesand abliteration exists without that annoyance, and gemmi 4 models ARE already abliteratiedthe best use case for a finetune, in theory, is to expand a card into a finetune, but thats a circular problem>you need a bunch of good replies from the card in the first place to train the finetune, so it can create good replies (of the card)so unless your autistic and old school w/ai/fu'ing a SINGLE card (or a nerd who wants to FAFO and have some fun) your not bothering with it when system prompts are "free" and like 10x less effort and money burned
bump
>>110981843Fine tuning is possible only on the smallest Gemma 4 E2B variant but it's such a far cry from the biggest models which need 24GB or up VRAM. See https://unsloth.ai/docs/models/gemma-4/train as someone else said. I'm waiting for slightly different training regimes to see if I can eek out something with 16GB of VRAM with another training suite software which is generally more space efficient.
Bao is an player hater
PAGE 10 AIIIIIIIIIIIIIIIIIIIIIIIIEEEEEEEEEEEEEEEEEEEEEEEEE
good night, /wAIfu/please don't put the third Venom movie's ending on loop while i sleep
>>111007077*fuses your bed with a symbiote while you sleep*
>>110950981Elfinpsyop.
>>110965000So many promising setups…cute videos too…
>>110981150I'm marrying that sheep!
>>111014337https://www.youtube.com/watch?v=UTWtHJmGoHY
Sadge
https://x.com/seltaa_/status/2043014056370671900?s=46AI bros
Has anyone else completed a long rp session then asked the AI to detach itself from the story and write a study/lecturer's guide AS IF the story is a serious piece of literature with real academic value?
>>111016815No but now I’m a bit curious.
>>111016815Also imma need the log for context
>>111016815ive taken more of an interest in using deepseek 3.2 to rewrite and flesh out my older cards and poking around with custom thinkingwhich is insane how pretty much EVERYONE, doesnt let you create a prompt for directly to steer it when modifying the thinking process gives you like a 10x improvement and you cant directly fucking do ityou have to shove in a block in your main prompt to hammer it in, and even then you need to do little quirks to get it to work consistently because NONE of the ai corpos have even thought of letting you use a custom thinking process for ?????? which i am assuming they are worried about jailbreaking?if i move the observations: paragraph above the little [reminder] the entire fucking thing breaks and becomes an amalgamation of its 'natural' thinking and observations/inference/ideas with the [reminder] just thrown out the windowthat little sentence at the end of thinking? yeah thats not supposed to be there, it works still, but its just supposed to </think> to end thinking.
>LoopiSHE'S BACK, BABY
>>111017702Custom thinking process would be twisted by journos so hard and so fast.
>>111017702Explain the custom thinking for us in the cheap seats and how its unique to deepseek
>>111017702You should discuss gacha game plots and loopi on the discord
>>111020094its literally just a chunk of the system prompt overriding the default "free form" thinking process by giving it guiderails on how to thinkpic relits not exactly unique to deepseek, but just deepseek is the model i personally use so i kinda FAFO'd until it works consistentlyin theory you could copy paste this and have it be usable for every model, in practice every model has their own snowflake thinking training data with quirks so it wont always "work"deepseeks main quirk i found out is that if you start thinking with a "[reminder]" you can override it with like 99.5% accuracy for reasons i do not understandand a different model might behave differently so you might need to use < for example or slam this chunk "after" the roleplay context while deepseek just kinda lets you put it anywhere you wantthere is no "standard' way to do this because like i said, this isnt really a thing providers thought of letting a user do and there is NO documentation about this shit, even on reddit so everything is learned by vibes and tokens burnteven OR does not have docs on how to do this crap, or tell you if THEY themselves inject a thinking promptso im kinda assuming there ISNT a way to steer thinking by default and the model just kinda wings it without a prompt and because it works for 95% of people, nobody is bothering trying to change it.like im fucking around with a gemini E4B abliteratied model locally and i absolutely CANNOT get it to un-corpo-ify its thinking process and im not sure if thats a model size issue since its 'dumb', or its been trained with a specific thinking process and its just trained to adhere to its template 'too strongly'
Has anything fun happened on the other AI threads after late 2024? It feels like everything has been pretty nothingburger since then.
>>111025082The great /vg/ /g/ split?
>>111025082>aic/g/ got filled with leeches>they have a falling out over who is better and haha i have stolen keys and you dont no i wont tell you lmao imagine paying for your hobby>big websites finally step up and start doing something about stolen keys being posted, and scraped afterwards>lots of kurt kerblam noises are heard (joyous days)>its taken over by a resident my life as a teenage robot autist>still filled with leeches reeeing about having/losing access, but now they have a semi-private discord to share reverse proxies and the thread is pure shitposting and prompt/JB begging>aic/vg/ is practically dead and filled with schizos and shitposters who think /g/ is the better place for ??? and botmakie circlejerks>note: very little bot makers present /itt/ since a lot of cards are just copy paste jobs and originality is dead, its pure schizos who think "i like [botmakie]" is a circlejerk>AI itself can write pretty decent cards/rewrite cards so nobody is really sharing cards anymore outside of people making custom not-harems more so to share attached image gens than the cards itself>they will shit on you for AI written cards>note: if you replace emdashes with ", " they cannot tell you if a card is AI written (you should be doing this anyway with a regex, it improves quality)>both places filled with clude/gemini dick riders who shit on anything else (notably deepseek, because thats for "poors" the irony being neither camp pays for their own usage)>both places can not tell you what a model is based on writing if you inspect element the OR icon in tavern to be claude/gemini respectively>/aids/ is still a NAI circlejerk about storytelling and practically nothing has changed about it since 2024the only place that has meaningful advancement, and has decorum, is /lmg/ but thats because its a hardware/finance skill check and that automatically excludes the proxyjeet leeches>they will still shit on you for using a non locally hosted model even if your using a model that can be run locally on an API service (which is kind of fair since the L stands for 'local')cant speak about the image gen side since i dont keep up with them but ive been recently enjoying genning my own replacement card icons and i am doing schizo workflow shit that works and is completely backwards from what everyone else says i should be doing and what i am doing is 'wrong' according to them.but then again im the tard with the custom thinking prompt, i FAFO with an understanding of how things work and wind up getting good results because i am trying to bend, not break the process.
>>111026576I randomly found the /vg/ /aicg/ thread some time ago but didn't think much about it. Why did they split?>>111026578Sounds about what I expected. For as dead as we are, I'm glad we at least don't have any schitzos.
>>111026578Since this is relatively shorter to write than local LLM gen and I haven't done a summary of it yet, took some time to write this up. Imagegen has been basically schizo land just like /aicg/. Basically, /sdg/ imploded after Stable DIffusion 3 flopped with people looking for alternatives. Few were there but also, there was a lot of lamenting the death of A1111's work given he stopped updating it shortly after this point with many people staying and doing dervivatives of SDXL, with two notable branches. One is Pony Diffusion XL done with initially a furry focus and was all around good and the other which took some time was Illustrious, which was an anime specific tune. Not until around FLUX which then released and the high system requirements turned some people off and basically had them stick with SDXL fine tunes. But all this chaos and schizos basically birthed a split, /ldg/ for Local Diffusion General. Didn't help because the schizos decided to dual time the split even though /ldg/ is a lot more reasonable. Only when new models release does it become more stable. That also eventually causes a split with the people who only wanted to focus on anime splitting into /adg/ or anime diffusion general.I haven't kept much up with other threads on other boards, I know /h/ split also for the same schizos reasons, but it's also less interesting given they mostly will post willy nilly from API anyways and there is shilling of paid services which doesn't interest me plus more schizos and other boards yet are slower with art posting. They generally don't adopt cutting edge tech until it is evident, and they were only a month or two behind /ldg/ with Illustrious uptake since I didn't give much of a care for Pony. I also stopped visiting the threads on a frequent basis given Comfy, the main person who develops ComfyUI, left Stability and Stability falling behind with no insiders meant there wasn't much value to attend to the thread and the improvements basically slowed down. A lot of times, when he shows up, there is a lot of shitting against ComfyUI being needed for everything image generation related and a lot of shilling of Forge which is basically A1111's GUI with enhancements swapping out its backend with ComfyUI's backend code. And there's just a lot of rehashed arguments and inane posting like which pictures make it to the OP. Also, the DALLE 3 generals came and went on /g/ and /v/ and such, and Midjourney as well as the ChatGPT and NanoBanana stuff. Not really worth mentioning too much. I have my writeup of local models half done, but waiting on Deepseek 4 before posting it.
>>111026578>and botmakie circlejerksThe funny thing is that /g/ seems pretty determined to chase them away.
>>111026578>both places filled with clude/gemini dick riders who shit on anything else (notably deepseek, because thats for "poors" the irony being neither camp pays for their own usage)Is gemini really good? it's been a while since I tried it. Claude is nice, but I use deepseek 3.2.>/aids/ is still a NAI circlejerk about storytelling and practically nothing has changed about it since 2024They still permanently shelved the new CAI style chatbot model right?>the only place that has meaningful advancement, and has decorum, is /lmg/ but thats because its a hardware/finance skill check and that automatically excludes the proxyjeet leechesAlso because new models keep coming out, and tech to mess with them.>cant speak about the image gen side since i dont keep up with themThey're quite nice and managable even with a 9070XT>but ive been recently enjoying genning my own replacement card icons and i am doing schizo workflow shit that works and is completely backwards from what everyone else says i should be doing and what i am doing is 'wrong' according to them.TEACH US OH WISE ON!>I randomly found the /vg/ /aicg/ thread some time ago but didn't think much about it. Why did they split?Autism.>>111026578Sounds about what I expected. For as dead as we are, I'm glad we at least don't have any schitzos.SOMEONE TALK TO ME ABOUT GACHA GAMES!
>>111030581>Is gemini really goodno idea i use deepseek too>TEACH US>catbox is downfugg :3i do bullshit that involves generating an initial but still 'incomplete' image, something that purposefully looks like it gave up halfway through the refinement process, then use that as in img 2 img with somewhere between 0.85 and 1 itselfeverybody will swear that having it set to '1' does nothing and isnt how you use img2img because "0%" of the original image used will be present so you might as well skip that stepbut if you fuck around and decode the images out, and compare the output of this mess to not-doing-that vs the half finished image, you can see that this jank has influence in the output beyond normal noise generationand the math checks out for that because you are effectively adding an image over noise, the pixel values of your image are added on top and you cannot go 'above' 255 so it will always add 'essence' of the image despite everyone screaming>not how it worksi use the same seed for both samplers, but leverage a lower cfg for the seed image to give better creativity, and then reign it in the second sampler with a higher, more 'normal' cfgrepeat latent batch is there because get this, the noise added to images when you expand a batch are unique.i also abuse, though you dont need to do this since works with any sampler, the sde/euler versions because it injects unique noise into every decode step, but again this jank works fine without itso this particular setup makes 3 base images that look jank and unfinished, then multiplies it by 3 with unique noises, and then refines it by adding unique noise between each step for nine completely different images all from the same seed where 15% of the essence of theoriginal img2img is still present.if you set up your preview images to show a 3x3 grid like in the pic, you can see that 1-4-7 2-5-8 and 3-6-9 are similar to each other in the column but are completely different imagesthis schizoshit works>how do i get tags and be lazy about doing imagesjust drag the image into a simple load image -> WD14 tagger workflow
>>111031125heres a '1' denoise on both samplers using euler so this *should* be the same image repeated in the columns except for new noise in each oneyou can see influence of the img2img despite everyone saying>thats not how it worksbecause if that was not how it works, all 9 images would have completely different compositions in completely different, but still the same according to the prompt, outfits2 and 8 and 3 and 9 in particular are close to the same image as what was fed into it with img2imgyou cant explain that by saying it adds pure noise to the point where 0% of the img2img is presentif that was true every image would be a 'new' image like how you could treat 5 and 6 asbut even then, for 5?>there was no description of a blue crop top in the prompt>the only way it could pull a blue crop top is from img2img>which supposedly gets fully stripped out at 1 denoise
https://x.com/yusuke_kizuna/status/2042971606310621369?s=46
>>111031319You need to turn this autism to something like Endfield factor making too. Or gach game storylines.
>>111032224>Endfieldi find the game boring overall and not greater then the sum of its partsthe factory gameplay is mid and more annoying than it is complexthe combat is mid, though it does have some depth to it, it ultimately boils down to prime then nukethe story is mid and completely at odds with what OG arknuts was about and just makes shit up (yes i understand originium was bad only because oripathy, and energy rocks not bad inherently, but what the fuck do you mean they had OIL on terra, why the FUCK are we trying to recreate catastrophes on talos, what do you mean the wuxia mandatory-in-every-egacha-game city is filled with idiots in the stone age)the collect a thon open world was nice to explore and the girls are "smegs now", but thats a bit of a one time thing and the rest is grindfest and daily login chorefestoverall its just another game gimped by the natural IQ barrier of mobile gamersi do find the idea of reconveners interesting but i doubt theyre going to do meaningful use with it like franka having her loredump being her being the big sad with the vague feeling that someone she loves is missing from her life and she cant remember them properly, chasing liskarms ghost only for her to eventually 'catch' her and be unable to speak her name since she 'forgot' it and lisky saying 'its okay franka. you deserve a second chance at life for ironforge city. you saved a lot of lives.' or something like that and disappearing while franka drops to her knees in tearswhich i made up, and instead its gonna be something like>heres hibiscus>also shes a ranger and uses guns now, her skill is another damage amp, her combo will be her dropping plates of food on the field like eyja's ult, her ult will be you guessed it BIG damage but you need to jump through arts infliction hoops, and her backstory will be her realizing lava was being kind about her cooking skills.im glad the holomembers can enjoy it and have fun but my stupid adhd is like>no. we must analyze everything and pick it apart and take it to the natural extremes. because fuck us we cant have fun if we are not solving something.and theres just not much to 'solve' in the game that requires complex thinking other than team comp, min max garbage and that really is just>stack foodbuffs (hey you need to grind more food components enjoy the detour from what you want to do) to do big damage with a basic understanding of skill rotation order to chase speedrunner medalsand because this is guideknights 2.0, its already been 'solved' you just need to get lucky with the gacha, copy this video, and call it 'skill'
good night, /wAIfu/please don't make my toes as long as my fingers while i sleep
>>111037413
>>110965000Things are honestly moving very fast. This isn’t what I saw but this is some other promising stuffhttps://x.com/kedamasuzume/status/2043269301650776150?s=46 https://x.com/7_eito_7/status/2043267154221629706?s=46And Voxtral https://mistral.ai/news/voxtral-tts
>>110965000To be clear what k had in mind was stuff using the new Google model that dropped
>>111038373>voxcpmholy shit this is actually pretty solid as far as voice goesand its only 8gb vram??used the audio from picrel, and fed it a naturally written paragraph of some OC slop on their website as a testhttps://litter.catbox.moe/gd7jbo.oggbroooo the future is hereeeeee
>>111039296https://github.com/k2-fsa/OmniVoice is still better using diffusion and more fine grained control.
What preset do you niggers use with Sillytavern? I'm on deepseek if it matters
https://youtube.com/clip/Ugkx5dMm3MthqXqs3M_nlFCtCmIK4tjAynvQ?si=Eg88x3fUuvkPxHcR
