Giovanni

@realrebelai I understand. That sounds frustrating indeed. I hope you can solve these problems. Good luck and thank you.

@gi_aguiar26 it bricked my comfy last night so i had to rework them. its going to be a little bit tbh. im not sure why they require the dependencies they do but it completely conflicts with multiple source dependencies. i gotta figure out what i did wrong or what they did wrong

currently working on bringing in the LongCat Video Avatar 1.5 model into comfy. i have an merged fp8 model ill give yall when its ready. give me a few hours and ill have it done.

@heydoughogan @ltx_model I'm specifically talking about LTX. Proprietary models (APIA) are irrelevant and unimportant to the open-source community. These improvements I mentioned are very necessary for us to have a consistent open-source model to compete with proprietary models.

This LoRA is essentially a post processing workflows, that’s what it unlocks. Use whatever model you need to generate your video or image! The future will be filled with hundreds if not thousands of models that are all good/bad at different things. LTX has filled a big gap in actual production with this LoRA and now makes high quality delivery beyond just resolution possible! It’s pretty great.

What can LTX HDR unlock for VFX workflows? @heydoughogan put together a deep dive covering setup, HDR output, Blender relighting, and why this matters beyond standard SDR AI video pipelines.

@HappyHorseATH We in the open-source community all want open weights, not this API crap.

👀 Hey there! Unlock your Perks HappyHorse officially launches, and we're not showing up empty-handed. 🐴 Join HappyHorse, let's get creative together✨ #HappyHorseAI #happyhorse #happyhorseEarlyBirdPrice

@ComfyUI @AlibabaGroup Since you haven't released the weights for the Happy Horse model, you've lost a lot of credibility in the open-source community.

ComfyUI now supports HappyHorse 1.0, Alibaba’s cinematic video generation model. Highlights: - Cinematic aesthetics, 1080p / 15s multi-shot consistency - Audio-visual sync: lip-sync dialogue, ambiance, and expressive vocals - Motion-preserving edits: transform footage or replace/insert subjects while keeping motion & composition

@AlibabaGroup Since you haven't released the weights for the Happy Horse model, you've lost a lot of credibility in the open-source community.

Honored to be recognized by TIME as one of the 10 Most Influential AI Companies of 2026! This recognition reflects our ongoing commitment to building a full-stack AI ecosystem rooted in open-source innovation. We’re encouraged to see the Qwen model series gaining traction well beyond China, thanks to the support of our global community.🚀

@AlibabaGroup All the hype surrounding the model was generated by the open-source community. We were all disappointed that it's a closed model (API). Please honor all of us in the open-source community and make available an open-source version (smaller and less powerful) for local use.

Discover HappyHorse’s advanced video capabilities in the full article: alizila.com/alibaba-rolls-…

We’re excited to roll out HappyHorse 1.0 in limited beta. It is now accessible to creators globally via the HappyHorse official website or through API service on Alibaba Cloud Model Studio.✨ Bring your stories to life with cinematic-grade video creation! #AlibabaAI #HappyHorse

@AlibabaGroup All the hype surrounding the model was generated by the open-source community. We were all disappointed that it's a closed model (API). Please honor all of us in the open-source community and make available an open-source version (smaller and less powerful) for local use.

@ltx_model We all want to know how the new data training for the new version of LTX is progressing, aiming to correct general human anatomy issues (body horror), more complex movements, and physics.

LTX HDR beta is now live. Every AI video model before this one output 8-bit SDR only. Fine for social clips. The format falls apart the moment you try to grade. Highlights clip. Shadows crush. AI footage won't composite cleanly against higher-bit-depth CGI. Resolution was never the real issue. Dynamic range was.

@ltx_model We all want to know how the new data training for the new version of LTX is progressing, aiming to correct general human anatomy issues (body horror), more complex movements, and physics.

We didn't ship HDR and ask if the output held up. We asked studios to break the tech. Gear Productions. Magnopus. Asteria. The result: an HDR IC-LoRA on LTX-2.3 producing 16-bit HDR EXR files. Gear ships LTX shots as final deliverables. Magnopus integrates AI assets onto LED walls. Asteria used the workflow on ChikaBOOM!, a Tribeca 2026 selection. The line between generated and filmed disappears. Their words, not ours. Check out the full story. ltx.io/model/model-bl…

@ltx_model Much more important than HDR is first fixing the basic problems of human anatomy (body horror), physics and consistency. For the next update, new training is needed in general human anatomy, physics and more complex actions. That's what we need for the next update!

@ltx_model Much more important than HDR is first fixing the basic problems of human anatomy (body horror), physics and consistency. For the next update, new training is needed in general human anatomy, physics and more complex actions. That's what we need for the next update!

Something big is coming. AI video looks good until you try to use it in production. Our latest research on HDR video generation redefines what pre-trained models can do. Stay tuned.

@EtherCoins @ltx_model Furthermore, and most importantly, new data training aimed at correcting the major problems of human anatomy (body horror, mutilated limbs, poor proportions and physics), especially in T2V. Thank you so much, you guys are amazing.

@ltx_model We need enhanced multimodal reference capabilities for stronger character/garments/style consistency a-la Seedance 2 - If you can achieve that we are golden LTX team.

@ltx_model @EtherCoins Furthermore, and most importantly, new data training aimed at correcting the major problems of human anatomy (body horror, mutilated limbs, poor proportions and physics), especially in T2V. Thank you so much, you guys are amazing.

@EtherCoins Will take this feedback to the team🫡

@ltx_model @haoailab Please perform a new data training to correct the major flaws in anatomy generation and blurring of human movements for the next LTX T2V update. There are still many problems with body horror, mutilated limbs, crooked hands and feet, and amputated legs. This is very frustrating.

Love seeing teams build on LTX 🙌 Great work from @haoailab

@Lightricks Please perform a new data training to correct the major flaws in anatomy generation and blurring of human movements for the next LTX T2V update. There are still many problems with body horror, mutilated limbs, crooked hands and feet, and amputated legs. This is very frustrating.

20 seconds of generated video. As in continuous. As in one prompt. Go try it out 👇

@ltx_model Please perform a new data training to correct the major flaws in anatomy generation and blurring of human movements for the next LTX T2V update. There are still many problems with body horror, mutilated limbs, crooked hands and feet, and amputated legs. This is very frustrating.

1,000,000 Hugging Face downloads, and counting! LTX-2 grew the way we believe products should: in the open, with the community, through real-world use. Grateful for everyone who helped get us here and excited for what’s next. Download and try it for yourself here: huggingface.co/Lightricks/LTX…

@ltx_model Please perform a new data training to correct the major flaws in anatomy generation and blurring of human movements for the next LTX T2V update. There are still many problems with body horror, mutilated limbs, crooked hands and feet, and amputated legs. This is very frustrating.

A faster model does not just save time. It makes experimenting easier. You can try more, adjust faster, and keep the momentum going while the idea is still fresh. LTX-2.3 vs Wan2.2 says it better than we can ⬇️

@ltx_model Please perform a new data training to correct the major flaws in anatomy generation and blurring of human movements for the next LTX T2V update. There are still many problems with body horror, mutilated limbs, crooked hands and feet, and amputated legs. This is very frustrating.