CryptoMcKenna

Phosphene 3.0 ships today. A free, local video creation suite for Apple Silicon. 100% Open source. Fully redesigned. Supports: Text and audio to video, Characters, image to video, LoRA training, FFLF and Extend and much more. Train unlimited characters now 🧵👇

Phosphene 3.0 is out. Hope you guys like it. Would really love to see what you do with it... 💕

Phosphene 2.0 is out: A Lot of improvements all across the board. I think you guys will love it. If I see the demand is there I may add add agentic capabilities in next release. For now, in Pinokio: Stop · Update · Start. Show some 💕 by sharing and follow @PhospheneAI

🚨ALERT 🚨 JSON Separatists have breached the Veo 3 Video Gen Lab! Allied with Structured Prompt Enforcers and JSON Schema Realists, they're imposing "curly brace sovereignty" on all AI video prompts. Demanding no more "token-wasting natural language chaos." Reports of YAML sympathizers smuggling in indents! #JSONPromptWars 🧵

🚨 Re-release: World Reacts to Iran Nuclear Attacks on Israel 🚨 One year ago, I made my first and only truly political video. Most of you haven’t seen it, since it came out before I revamped Bizarro’s account. But it’s just as true today as it was then. 🖐️I expect some of you will unfollow Bizarro after watching it. So be it. Farewell, my bros. This was my raw response to the October 7th Massacre in Israel. And in case you're wondering: yes, Bizarro is a proud Israeli. 🫂 I truly hope the Iranian people are finally free from the tyranny of ayatollah oppression.

🌞GM brothers. Bizarro is back in action. I remade my Cinematron LoRA with Wan. Text-to-video Wan with LoRA is terrible, and I dislike the image-to-video outputs as well. Both consume too many resources. What am I missing? Hunyuan isn't production-ready, but I can get nice clips in 500 seconds and combine LoRAs as desired. as of now most optimal workflow for me is: 4o + paid tools to animate.

🚨 Testing, testing. Does Bizarro still have any reach at all? 🦵 Also, what's up with the third leg?

@Dentoshi getting close to the place everybody an their mother want to buy. we either didn't get it or we visit 50k.

And the worst monthly chart award goes to... gm?

👨‍🏭 It is, in fact, way harder to work with Hunyuan locally than using closed-source tools. 👨‍🏭 🤯 But the customization is off the charts, and we don’t even have the LoRAs we need yet—but they’re slowly appearing (camera movement/character motion) 🧨 I predict that in 6 months, 90% of my generations will be done locally. For my new video, it will be around 70%.

This really looks like final capitulation in alts, the fact that t was this quick is good news in my opinion. I would be more worried of a slow bleed.

🚨There is no moat in video generation as well🚨 Did you notice how good all the closed models have become? Luma Ray 2, Pika 2.1, Kling Elements, Pixverse v3.5, Runway, they’ve all hit high-quality, realistic video generation at impressive speeds. Some did it cheaper than others, but none of this required billions. Now we have @TXHunyuan, a very capable system. This is just the beginning, but I’m calling it now: We’re going to see full HD generations highly optimized to run on consumer cards before the end of the year. Expect "elements," image-to-video workflows, and easy-to-train LoRAs to become standard. If Hunyuan stumbles, the next contender is @LTXStudio. They’ve nailed speed and optimization but are still catching up on quality. Also probably 1 or 2 of these companies will survive as they will be selling inference rather than exclusivity of the models. For reference: This image was generated on my 3090 using my Steampunk LoRA in just 13 minutes. 848x480 93 frames 50 steps. Upscaled in Topaz.

GM, Bizarro has been doing a lot of testing and lora training but very little of video creation. It is what it is 😕 Hopefully we change that trend this week.

GM, fellow creators! 🚨 Bizarro has crafted the ultimate steampunk-style Hunyuan LoRA! 🚨 Should I release it for the community? 🤔 Here are the first 4 examples straight out of the forge, no cherry-picking. Your thoughts?

GM finally solved @FAL issue with LoRAs and now they work locally. 🚨 Releasing my Hunyuan Tesla Optimus LoRA 🚨 It’s here for everyone in the community to use and push their creativity. The aesthetics are very Flux-like, and the results? They speak for themselves. If you create something cool with it, I’d love to see what you come up with. civitai.com/models/1176055

@PeterLBrandt 10k on low conviction. 20k on very high conviction. more than 4 stops consecutive take a few weeks off. and reassess.

Please answer this question: If you have a $1,000,000 trading account, what's the maximum you are willing to lose on your next trade? I'll be interested in the avg answer by Verified vs Troll respondents. I will provide my answer tomorrow.

🚨 I’m starting a new project, and it’s going to be different in character. 🚨 🐒There’s probably not going to be any traction, and I’m done with the monkey business of trying to go viral. 🤿I’m diving deep into Bizarro' s brain, searching for the real stuff. I am not afraid to fail. 35s of voiceover in the bag. LFG!

Let’s talk Hunyuan. This LoRA was meant for my Bizarro character, but it turned out to be a steampunk powerhouse, capable of generating a cinematic atmosphere very faithful to the style of my generations. I used @FAL 's LoRA trainer on some of my Bizarro character Flux generations and also generated the videos in FAL. It’s pricey ($5 per LoRA, 20¢ per generation) but the results are pretty solid. We are getting there.👊 Only caveat. I am unable to get this quality on my 3090 at the moment with mentioned Lora.

NSFW AI gen people go hard. They not only have the best quality local generations on civit but are actually pushing the whole space forwards. Look a this dude, he wrote freking whitepaper about how to get Hunyuan to represent boobs and genitalia properly. reddit.com/r/StableDiffus… Whenever you need the best local workflows, you know where to search. Also NSFW video is advancing so fast people don't even understand what the implications are. Our kids will grow in a very challenging environment where people don't have any power of what others do with their likeness. it is extremely worrying to be honest.

[FOR ALL MY VIDEOS, ⬆️ CHECK ⬆️ HIGHLIGHTS] Bizarro Dimension: #45@SD=68SA 💊 When the world succumbs to the most absurd syndrome ever Your support means the world. Consider sharing if you enjoy my content ❤️.