PhD Student @ Hao AI Lab, Intern @ Anyscale
Joined December 2018
- Tweets 85
- Following 139
- Followers 78
- Likes 742
Photos and videos
Will Lin retweeted
May 27
Open source moves the whole field forward. Love seeing @haoailab bring real-time video directing to life with LTX-2.
May 27
🚀Generate a 30-second 1080p video in just 7 seconds!
We’re open-sourcing FastVideo Dreamverse: real-time vibe directing for video generation on a single NVIDIA B200 GPU with LTX-2 model @ltx_model
Repo: github.com/hao-ai-lab/FastVi…
Blog: haoailab.com/blogs/fastvideo…
3
17
152
12,613
You should read this thread.
It used to take about 25 seconds to generate a 5-second video on 8 Blackwell GPUs. The legends at @haoailab brought that down to just 4.2 seconds on a single Blackwell GPU… and then open sourced the tech behind it.
May 27
🚀Generate a 30-second 1080p video in just 7 seconds!
We’re open-sourcing FastVideo Dreamverse: real-time vibe directing for video generation on a single NVIDIA B200 GPU with LTX-2 model @ltx_model
Repo: github.com/hao-ai-lab/FastVi…
Blog: haoailab.com/blogs/fastvideo…
39
182
1,722
170,676
Will Lin retweeted
May 27
🚀Generate a 30-second 1080p video in just 7 seconds!
We’re open-sourcing FastVideo Dreamverse: real-time vibe directing for video generation on a single NVIDIA B200 GPU with LTX-2 model @ltx_model
Repo: github.com/hao-ai-lab/FastVi…
Blog: haoailab.com/blogs/fastvideo…
33
115
730
223,238
Will Lin retweeted
May 1
Such an honor to share that our 2016 paper GeePS just received the EuroSys Test-of-Time Award 🫡🚀🏆
It was actually my first system paper (and obviously my first MLSys paper, too) in phd -- and arguably the first paper to systematically tackle GPU memory swapping for deep learning, right after AlexNet moved DL onto GPUs.
It has been 10 years! The ideas are everywhere. A short thread on what we did and where it went 🧵
3
8
81
9,363
Will Lin retweeted
Apr 20
So excited to send our @wlsaidhi to present in this Dynamo event during GTC -- we shared a lot of latest advances, knowledge, OSS development in FastVideo to the greater Nvidia / Dynamo community. Our dreamverse and fastvideo projects get so many excitement from the community.
Let's keep the FastVideo Dynamo collaboration going and push for real-time videogen in 2026!!!
What a night for the NVIDIA Dynamo community 📸
From OSS commits to production-scale inference, NVIDIA Dynamo is quickly becoming part of the modern AI stack enhancing stability, reliability, and speed.
One thing was clear: faster inference is the new frontier—and Dynamo is the skill to master. Inference is hard (really hard), and the game is shifting.
And the conversations didn’t stop there. An incredible turnout stayed on to connect, share ideas, and celebrate the growing inference ecosystem.
Huge thank you to our speakers from @alibaba_cloud, @baseten, @haoailab, @intel, @Pinterest, @PrimeIntellect, and to everyone who joined us and continues to push Dynamo forward.
📗 nvda.ws/4mhXaW1
1
16
1,574
Will Lin retweeted
Apr 9
(1/5) FP4 hardware is here, but 4-bit attention still kills model quality, blocking true end-to-end FP4 serving.
To fix that, we propose Attn-QAT, the first systematic study of quantization-aware training for attention.
The result: FP4 attention quality is comparable to BF16 attention with 1.1x–1.5x higher throughput than SageAttention3 on an RTX 5090 and 1.39x speedup over FlashAttention-4 on a B200.
Blog: haoailab.com/blogs/attn-qat/
Code: github.com/hao-ai-lab/FastVi…
Checkpoints: huggingface.co/FastVideo/14B…
9
31
244
37,354
Will Lin retweeted
What a night for the NVIDIA Dynamo community 📸
From OSS commits to production-scale inference, NVIDIA Dynamo is quickly becoming part of the modern AI stack enhancing stability, reliability, and speed.
One thing was clear: faster inference is the new frontier—and Dynamo is the skill to master. Inference is hard (really hard), and the game is shifting.
And the conversations didn’t stop there. An incredible turnout stayed on to connect, share ideas, and celebrate the growing inference ecosystem.
Huge thank you to our speakers from @alibaba_cloud, @baseten, @haoailab, @intel, @Pinterest, @PrimeIntellect, and to everyone who joined us and continues to push Dynamo forward.
📗 nvda.ws/4mhXaW1
5
14
71
8,690
Will Lin retweeted
Mar 30
lol maybe pip install fastvideo a little earlier 😂
just kidding. This shows that making videogen real-time and cheap is so important and already hard enough, and turning it into a real product is even harder. we’re still working hard on our side too.
Mar 30
Why OpenAI really shut down Sora techcrunch.com/2026/03/29/wh…
1
2
24
5,814
Will Lin retweeted
Mar 24
Hot take: the metaverse was never a bad idea.
It was just too early, and built with the wrong stack.
The breakthrough version of the metaverse won’t be a manually designed world with heavy headsets. It should be a world generated on demand and responsive to what the user want to see, do, create, and feel.
Imagine environments that reshape instantly from language, mood, memory, collaboration, or story. That is a totally different product experience from traditional rendering.
So, real-time video diffusion is one of the key unlocks for that future here. Once generation is faster than consumption, we are no longer exploring a static world, but you are basically live- and vibe-directing it.
That’s a big reason why I’m spending so much time on real-time diffusion / FastVideo right now. Super bullish on this direction
3
9
77
14,252
Will Lin retweeted
Mar 19
Love seeing teams build on LTX 🙌 Great work from @haoailab
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
5
5
42
6,314
Will Lin retweeted
Mar 19
🚨 @soumithchintala followed @haoailab
Hao AI Lab (@haoailab) is a UCSD lab working on strong, efficient and scalable ML systems, with projects like DistServe and FastVideo getting traction beyond academia.
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
1
9
3,775
Will Lin retweeted
Mar 18
❌Metaverse
✅Dreamverse
Generate and edit 30s 1080p videos (with audio) faster than you can watch them: dreamverse.fastvideo.org
4
13
3,290
Will Lin retweeted
Mar 18
@NVIDIAAI Please consider giving us Rubin; we'll give you real-time 4K infinite world model gen 🤔
Mar 18
Wow! The Vera Rubin demo looks great but real-time editing is actually already here on a single B200!
Try Dreamverse today and generate 30s 1080p videos (with audio) faster than you can watch them.
Demo: dreamverse.fastvideo.org/
2
23
1,725
Researchers at Hao AI Lab (UCSD) have unveiled Dreamverse, a prototype interface powered by their FastVideo inference stack that fundamentally changes how creators interact with AI video generation.
FastVideo has optimized its inference stack to generate a 5-second, 1080p video clip in roughly 4.55 seconds on a single GPU. Because the generation is actually faster than the playback time, it enables a continuous, real-time creative loop.
Rather than writing one massive, highly engineered prompt and waiting minutes for a result (like with OpenAI's Sora), Dreamverse allows users to "vibe direct." Creators can use a chat interface to issue rapid, natural language adjustments like "change the camera angle," "slow down the motion," or "continue the scene" while the video is actively unfolding.
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
4
15
44
6,436
Will Lin retweeted
Mar 18
Wow! The Vera Rubin demo looks great but real-time editing is actually already here on a single B200!
Try Dreamverse today and generate 30s 1080p videos (with audio) faster than you can watch them.
Demo: dreamverse.fastvideo.org/
A breakthrough in real-time video generation.
As a research preview developed with @NVIDIA and shared at @NVIDIAGTC this week, we trained a new real-time video model running on Vera Rubin. HD videos generate instantly, with time-to-first-frame under 100ms. Unlocking an entirely new creative paradigm and bolstering the foundations of our General World Model, GWM-1.
Real-time generation opens a fundamentally different design space for video models and world simulation. We're investing in co-designing our models alongside advances in hardware to keep pushing this frontier.
12
28
5,859
Will Lin retweeted
Mar 17
new playground from FastVideo.
Addictive poison: endless scene tweaks for no reason. Just because it can regenerate the whole video in a second.
trippy hallucinations. Session over, I surrender
dreamverse.fastvideo.org/
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
1
7
31
4,998
truly a huge breakthrough for video gen, got to play around with the model the past couple of days and felt like the next leap forward in terms of actually steering videos. and it’s SO fast. congrats @matthewnoto73 and team!!
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
2
15
3,541
Will Lin retweeted
Mar 17
Dreamverse is our AI video engine that generates a video faster than you can watch it.
30s of 1080p video in 4.5 seconds. One GPU. Real-time editing.
This is vibe-directing.
dreamverse.fastvideo.org
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
12
14
42
6,786
Will Lin retweeted
Mar 17
@haoailab and Fastvideo are out with an awesome real time prototype.
Getting us one step closer to Video as the new UI.
Mar 17
So excited that our prototype Dreamverse is finally out!!
I increasingly believe real-time diffusion is not just about latency optimization. It might unlock a fundamentally new interface for creating and using AI video. Once high-quality videogen becomes faster than playback, you no longer have to write one giant prompt, wait forever, and then start over. You can stay in the loop: generate, watch, chat, revise, continue. Or maybe, in the future, you can just interact with the world like how humans do it. 😯😯
The key point here is videogen stops feeling like submitting a job and starts feeling like directing. That is the core idea behind Dreamverse, and what we call vibe directing. Check an early prototype from us: dreamverse.fastvideo.org/
I also really want to shout out our FastVideo team here.
- We got 5s 1080p video audio generation down to real time in less than a month 😉
- we then turned that systems capability into this vibe-directing prototype experience in about a week 🫡.
That is an extremely high bar of execution. Small team, strong taste, and the ability to move from infra breakthrough to prototype demo very fast. I am genuinely very proud of what the team built.
Please try the demo and let us know what you think 👇👇
1
6
206
Will Lin retweeted
Mar 17
So excited that our prototype Dreamverse is finally out!!
I increasingly believe real-time diffusion is not just about latency optimization. It might unlock a fundamentally new interface for creating and using AI video. Once high-quality videogen becomes faster than playback, you no longer have to write one giant prompt, wait forever, and then start over. You can stay in the loop: generate, watch, chat, revise, continue. Or maybe, in the future, you can just interact with the world like how humans do it. 😯😯
The key point here is videogen stops feeling like submitting a job and starts feeling like directing. That is the core idea behind Dreamverse, and what we call vibe directing. Check an early prototype from us: dreamverse.fastvideo.org/
I also really want to shout out our FastVideo team here.
- We got 5s 1080p video audio generation down to real time in less than a month 😉
- we then turned that systems capability into this vibe-directing prototype experience in about a week 🫡.
That is an extremely high bar of execution. Small team, strong taste, and the ability to move from infra breakthrough to prototype demo very fast. I am genuinely very proud of what the team built.
Please try the demo and let us know what you think 👇👇
Mar 17
(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU.
Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live.
🕹️Demo: dreamverse.fastvideo.org
📑 Blog: haoailab.com/blogs/dreamvers…
Welcome to the era of vibe-directing 👇
3
8
47
5,980