Fake (@FakeUGC) header
Fake (@FakeUGC) profile picture

scaling ai models | https://teapaw.org/t.co/jdpW90wPWQ

Joined March 2022
Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)Photo by Fake (@FakeUGC)
We've partnered with @Microsoft to bring Hermes Agent to the enterprise. Hermes Enterprise now works inside Microsoft Teams, the Microsoft Store and Azure. In Teams, employees hand Hermes work like financial analysis, document drafts or inbox triage, and review the result in the same chat. Agent workloads run in isolated Azure Container Apps sandboxes. When someone finds a good way to do a recurring task, Hermes saves it as a skill. With Collective Wisdom, they can share that skill with their team after review, so the next person starts out with the best approach anyone has found to accomplish that task. The skills then grow along with the company. Companies with strict data residency or procurement rules can run Hermes Enterprise in their own Azure tenant. Our CTO Jeffrey Quesnelle presents how Hermes uses these sandboxes at Microsoft Ignite today, in session BRK448. Thanks to @M12vc and the @Azure Container Apps team. nousresearch.com/hermes-ente…
486
814
364
8,663
710,877
I'm so sorry Google. I didn't recognize your game. To be fair, you did not indicate at any point that you still had game like this. I'll take the L either way. I'm just pumped that we finally have a good Gemini model.
353
311
101
13,363
598,598
Fake retweeted
And here we go…GPT-6.2 Astra and GPT-6.2 Sol are yours to make more games with. Oh, and they’re also super efficient and really good at making slides, building websites, and tackling the work you were doing before you started making games. Rolling out today to ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users. openai.com/index/introducing…
🤖 Made with AI
585
1,184
450
13,128
971,895
Fake retweeted
Introducing Claude Meetings: add Claude to your team’s video call and it builds docs, whiteboards, and other artifacts with you while you’re still talking. Starting to roll out in beta on Team and Enterprise plans.
2,386
5,181
5,441
81,443
10,275,008
I can test a dance reel for my AI girl without organising a shoot or performing on camera. But a dance getting views isn't the same as her Fanvue making sales. First, I check whether the opening holds attention and brings people to her Instagram profile. Next, I make sure her other posts feel like the same character, aimed at older US guys. Then, I track Fanvue subscriptions and extra photo/video sets bought by visitors. I'd get that path working on Instagram before giving myself a second feed to fill on TikTok.
1
1
1
276
Fake retweeted
ONE image is all it takes to keep your ai model consistent every single time the tool doesn't remember her from last time. so every new picture or video it makes of her starts from scratch one board of her goes in every time > her face straight on close up > her face at three-quarter > full body from the front > full body from the back only the angle changes between the four. the body matters as much as the face you hand it that board each time so it has her whole look to pull from instead of inventing her BEST way to keep her consistent here's an example
6
13
269
14,496
Fake retweeted
people are PRINTING on reels with fake omegle calls right now a girl on omegle, a guy sees her face and loses it. this exact one got 38 million views and 504K likes it's one of the lowest effort formats you can make. only her half gets generated, the guy is a reaction clip from another video edited in 1. screenshot the girl from an omegle reaction clip that's already working 2. ask claude for a prompt that matches her look with your ai model in it 3. paste that prompt into nano banana with your ai model's image for the face and the screenshot so it copies the vibe of the shot 4. animate her in veo 3.1, then stack her under the reaction clip in capcut so it looks like an omegle call pick a reaction clip that's still picking up views and run these steps on it tonight
2
1
16
15,272
Fake retweeted
this is easily the easiest viral format for ai model content on reels right now... a stranger stops a girl on the street and asks about her outfit the mistake is prompting it like a shoot. clean camera, her talking straight to the lens, a perfect line. that's what makes it look staged a real one gets filmed by some random guy on his phone it shakes, it blurs, she walks out of frame and cuts him off mid question so that's what goes in the prompt: - handheld, whip-pan away, motion blur, she walks out of frame - every beat timestamped in whole seconds, 0-3s, 3-6s - her lines in { } with the language and accent named first - the question in brackets as an off-camera voice keep your ai model and the format the same every episode and only change the outfit
2
2
24
5,735
429
226
171
5,119
419,400
Fake retweeted
i haven't had a single original idea for my ai model in months.. every reel on her page started as somebody else's reel i only pull from 12 accounts whose reels already get views. if the format works for them it works for me the pick is always a real girl. never an ai page. an ai reel moves like nothing alive and that carries straight into yours i look for a clip where people are still reacting and no other page has used the format yet if every page already runs it i move on a real clip carries rhythm and movement a prompt can't describe. the way she shifts her weight, how her head moves while she talks the whole pick takes about 10 minutes. then the reel gets recreated with my model in it open one account from your follow list tonight and find one reel that's still picking up. if nobody else has run it yet it's yours
4
3
25
3,237
We made the first AI model you can have a realistic conversation with that was trained entirely on speech! Dessa is a full-duplex speech-to-speech model that you can run on consumer GPUs. This is a GPT3 moment for speech.
1,026
1,256
1,219
17,388
4,664,371
Fake retweeted
if your agency runs a real girl, feed her last photoshoot to ai and watch her page keep posting between shoots sooner or later every agency runs its real girl the way an ai model page runs. she stays the girl, ai just gives you another way to hit traffic and it's a lot easier the ones still shooting everything by hand are riding horses while the cars go past 20 photos of her face. bin any where the face shifts between angles, plastic skin, too symmetrical. log every keeper then build the same set for her body from there ai generates new outfits, new locations, new angles from her set. one real shoot keeps her onlyfans posting long after the camera is packed away paid video is $0.23 a second, so i only pay for the clips that need it. the rest run on free local models generate her next outfit from the set tonight and post it next to her real shoot
1
1
1
719
Fake retweeted
post a bikini photo in your first few weeks and watch instagram take your whole page down. it doesn't even have to be nsfw. an ai model page can get flagged for a bikini alone and the whole page gets pulled so nothing suggestive goes up until day 21. lip syncs, a coffee run, normal stuff only the "link me" in your bio makes it worse. people click it, see you've got fanvue linked, and report the page move the link out of your bio and put it inside a story highlight. that's how you get around it if the page goes down it's a new sim and day one again, so set the backup account up on a burner phone now before you need it
2
10
808
Fake retweeted
f*ck it.. here's the exact structure i break every viral reel into before i remake it: give gemini any clip with this prompt: "describe the video frame by frame, scene by scene, with an accurate description of the story and what happened" the breakdown only works if every timestamped block fills the same fields: 1. the shot. camera move and distance, "wide POV, eye level, handheld walk" 2. performance. body language and energy, "sits still" 3. dialogue with delivery. the stammer written out word for word 4. sound. the specific noise in that block, footsteps then the loud metal crash of the chain-link fence 5. her look. repeat it every block, black graphic top, ripped black tights, grey furry boots, silver pendant 6. find the turn. gemini names the exact second the story flips, 0:08 she stands and backs him into the fence gemini writes a video prompt at the end with [MY MODEL] as a placeholder push that prompt into seedance with your ai model as the reference, or a scene board of her you made in chatgpt. keep the turn beat where it is
1
1
8
825
Fake retweeted
Opus 5.5 >>>>>> Gemini 4 Argon.
55
14
2
371
10,698
Fake retweeted
Introducing Copy. Any pose you find, recreated with your character. Drop the photo in, and it's yours. Live now, go try for free!
16
30
8
356
112,293
Fake retweeted
it's not crazy to think that a lot of the content we see in our IG feeds is made in this exact way: > autonomous agent scrapes content that's performing > extracts why it works > rewrites the script / structure > uses Higgsfield API to create the video > populates the timeline with 1000's of similar videos to spread a narrative
🤝 Paid partnership
Introducing OpenAI’s Dots x Higgsfield. Your always-on Higgsfield creative crew keeps working while you’re away. Check in by text, call or email, and pause the work whenever you need. Powered by GPT-6.1 Sol.
9
5
2
204
21,968
Fake retweeted
FOUND AN UNCENSORED LOCAL VIDEO MODEL TO CREATE ADULT CONTENT 👀 NSFW Wan 1.3B was fine-tuned specifically for adult content • text to video • based on Wan 1.3B • trained on NSFW data • around 2.8GB for the newer checkpoint • can run locally don't download the old e10 checkpoint though the repo now recommends: wan_1.3B_exp_e14.safetensors check it out: huggingface.co/NSFW-API/NSFW… test it locally with fictional adult characters
SOMEONE JUST TURNED QWEN3.5-9B INTO AN UNCENSORED LOCAL MODEL and it runs on your own laptop • 5.2 GB Q4_K_M • 18/18 on the best refusal test • 262K context • Ollama, llama.cpp and MLX • vision support on GGUF the wild part is how small it is download it here: huggingface.co/lukey03/Qwen3…
86
881
21
15,001
1,531,548
Fake retweeted
i make my ai model uglier on purpose.. just human. one eye a bit smaller than the other. a shadow under each one. loose hair. one mole. visible pores those five things are typed into the prompt. the generator will default to flawless unless told otherwise name a phone camera instead of a studio lens. the softness is what makes her look real a 10 out of 10 face makes him scroll. too perfect. he wouldn't bother sending a message a 7 posting selfies from a kitchen he recognises gets a message by day three. by the end of the week he's sending paragraphs she didn't ask for the tell is the skin. no pores, no texture, clocked as fake in under two seconds i run over 20 versions of every shot and cut about 80%. first thing i check is the symmetry the pretty ones get saved to a folder. the ones with frizz get posted
1
1
8
780
Fake retweeted
nobody has built an ai model that flirts in german yet.. 100 million people speak it. germany, austria, switzerland. and every girl they're shown was built to speak english flirting in a second language is homework. he's translating every line in his head before he answers. in german he just answers so when a girl from leipzig texts him that her train is late again, in his own language, he stops feeling like a customer every course in this niche is taught in english. every student builds her in english. every one of them ends up fighting over the same guy in ohio in english she's one more girl he scrolls past. in german she's the only one the voice clone that makes her talk speaks 74 languages for the same $22. the dms run through translation, so the chatter doesn't need a word of german no one is there to undercut her either, so she can charge more than the english pages pick the language before you pick the face comment "DEUTSCH" and i'll send you the niche selection section of my system (must be following + RT)
3
7
1,049