Hacker News Reader

Comment by xnx | Hacker News Reader

This looks great, but I'm confused by this part:

> Veo sample duration is 8s, VideoGen’s sample duration is 10s, and other models' durations are 5s. We show the full video duration to raters.

Could the positive result for Veo 2 mean the raters like longer videos? Why not trim Veo 2's output to 5s for a better controlled test?

I'm not surprised this isn't open to the public by Google yet, there's a huge amount of volunteer red-teaming to be done by the public on other services like hailuoai.video yet.

P.S. The skate tricks in the final video are delightfully insane.

View on HN

> I'm not surprised this isn't open to the public by Google yet,

Closed models aren't going to matter in the long run. Hunyuan and LTX both run on consumer hardware and produce videos similar in quality to Sora Turbo, yet you can train them and prompt them on anything. They fit into the open source ecosystem which makes building plugins and controls super easy.

Video is going to play out in a way that resembles images. Stable Diffusion and Flux like players will win. There might be room for one or two Midjourney-type players, but by and large the most activity happens in the open ecosystem.

> Hunyuan and LTX both run on consumer hardware

Are there other versions than the official?

> An NVIDIA GPU with CUDA support is required. > Recommended: We recommend using a GPU with 80GB of memory for better generation quality.

https://github.com/Tencent/HunyuanVideo

> I am getting CUDA out of memory on an Nvidia L4 with 24 GB of VRAM, even after using the bfloat16 optimization.

https://github.com/Lightricks/LTX-Video/issues/64

Yes. Lots of folks on reddit running it on 24gb cards.

Yes you can, with some limitations

https://github.com/Tencent/HunyuanVideo/issues/109

Stable Diffusion and Flux did not win though. Midjourney and chatGPT won.

“Won” what exactly? I have no issues running stable diffusion locally.

Since Llama3.3 came out it is my first stop for coding questions, and I’m only using closed models when llama3.3 has trouble.

I think it’s fairly clear that between open weights and LLMs plateauing, the game will be who can build what on top of largely equivalent base models.

The quality for SD is no where near the clear leaders.

> The quality for SD is no where near the clear leaders.

It absolutely is. Moreover, the tools built on top of SD (and now Flux) are superior to any commercial vertical.

The second-place companies and research labs will continue to release their models as open source, which will cause further atrophy to the value of building a foundation model. Value will accrue in the product, as has always been the case.

SD will also generate what I tell it, unlike the corporate models that have all kinds of “safeguards”.

You must be stuck at SDXL for posting something absolutely and verifiably false as the sentence above.

I wonder if the more decisive aspect is the data, not the model. Will closed data win over open data?

With the YouTube corpus at their disposal, I don't see how anyone can beat Google for AI video generation.

Reply on news.ycombinator.com