تدريب Shadowing: SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer, [ICLR 2026, Oral] - تعلم التحدث بالإنجليزية عبر الفيديو

اختر وضع التدريب

كل وضع يفتح غرفة تدريب بملء الشاشة.

125 جمل

إذا كانت الجمل قصيرة أو طويلة جدًا، انقر على Edit لتعديلها.

Thank you, Luke, for the amazing opening.

So good morning and thanks for having me.

My name is Jun Song.

I'm from the University of Hong Kong and NVIDIA.

It's really my great honor to present this work, SANA Video, for our SANA team.

So currently, generative AI is evolving rapidly with model sizes growing larger and larger.

At At the same time, AVIDIA GPUs are becoming increasingly powerful.

We can foresee that the future visual generative models will demand even more computational resources.

So actually, generating images and videos are significantly more computationally demanding than processing just text.

For example, the inference for Flux or CogVideoX exceed the Lama 3,

and an order of magnitude higher than the VIT models.

So they are restricted to cloud deployment and remain impractical for consumer GPUs or age devices.

So actually, efficiency is a forever topic.

So far, we have explored several directions for efficient video generation.

First, in Sanaa, we tried to reduce input tokens by using our self -developed deep compression autoencoder, DCAE.

We also try to use linear tension to replace Softmax 1, enhance the architectural efficiency of the model.

And we also already explored inference step distillation in Sanaa's spring project.

The fewer the inference steps, the faster the speed.

So the last part is what we are going to talk about today, the hybrid AR plus diffusion for long video generation.

Actually, we can generate video block by block causally and diffuse within the block.

So video generation requires much more computation than images.

SANA can generate one image within just one second.

However, generating a five -second video with one model actually takes nearly 30 minutes.

So to boost the video generation efficiency, we focus on three key technical areas, the high ratio VAEs and leaning attention mechanisms,

especially optimized for video tasks.

Crucially, the block KV cache based on leaning attention enables the generation of much longer videos.

So to get a video generation model, actually, we have two ways to go to try to choose.

First, we can just train from scratch.

But only when we have hundreds or thousands of GPUs,

like H100 GPUs, and it will take you almost several months to train one model.

And it's, of course, not affordable.

Or actually, we can start from a semantic aligned text -to -image model.

Here we start from our SANA 1 .5, which is a 1 .6 billion parameter text -to -image model trained on our DCAE.

Actually, since our video model is based on image model, so the linear tension DIT actually is almost the same as our SANA,

but we're introducing two modules for temporal modeling, and we will talk about them later.

Finally, on the left at the top, we implement block causal linear tension module to achieve constant kv cache memory.

So for any length video generation, So first, let's talk about linear attention in video generation model.

Actually, in video tasks, linear attention provides even more significant acceleration than in images,

as the video length and resolution are much larger.

Our results show that for long videos, SANA video is 1 .5 times faster than the one model.

And additionally, at high resolution, our linear attention is four times faster than the flash attention.

And these improvements make a long form, high resolution video generation much more practical.

However, we find that the linear attention map is far different from the softmax attention.

As you can see in this figure, the linear attention is super dense, while the softmax attention is rather sparse.

And this demonstrates that the local contacts captured by the linear tension is actually insufficient.

To enhance the temporal modeling, we introduced two key improvements.

First is the 3D rope for the linear tension and temporal convolutions.

And both are highlighted in the orange box.

While 3D rope doesn't solve the tension map sparsity problem entirely, it do help improve attention low quality.

Combined with our spatial temporal mix FFN, it significantly boosts the overall performance.

Here we have the training loss.

The green line stands for the 3D rope and the additional temporal convolution.

And to further support long video generation, we leverage the unique properties of linear attention to implement an efficient block KV cache.

So let's quickly review how this works.

In linear attention, we call it the attention map, or the state.

Actually, it has a fixed shape of D times D, regardless of the input sequence length.

So when processing the first block here, we obtain the state, S1.

And for the second block, the state is updated additably,

meaning that we only need to add the current block's contribution to the previous cache.

And generally for any block I,

actually the current state is updated as adding the current block contribution the KV here to the previous cache to ask.

So now let's see how we enable long video generation with block of causal linear attention.

For the first block here, the green one, we compute the state as one, shown in the green box.

After we compute first block the state, we cache it, it's shown in the red box.

Then we calculate the KV for the next block and add it to the previous one.

Here we take the cached state from block one in the red box and updates it into the block two, which is in the green color.

And the subsequent process follows the same logic.

The new state is cached, and once the next block competition is finished, they are summed together. And so on.

So, token in Valenian attention will become larger, more and more as the video length become larger and larger.

While our causal linear tension maintains a constant KV cache potentially make minute length video generation possible.

So here in conclusion, we can have two advantages of block linear tension.

First, we have constant memory usage regardless of the input sequence length and second we have the global history context.

So one last question is how to achieve causal FFN with temporal conf.

So let's take two block situation as example.

During causal training the block one cannot foresee block two so we pad a zero token here between the two blocks.

However block two must see block one's token so that it's correct the causal modeling.

So here we catch the last token from block one as the first token in block two.

So finally the two blocks are separated by the zero padding token

which is subsequently removed to to avoid affecting the final computation.

And the block two used the last token from block one making the the causal modeling available.

So here's some results, compared to the current state -of -the -art video models, Sonar Video performs competitively on VBench,

and it takes only 18 seconds to generate a 1K resolution video on a single H100 GPU.

Here we show some examples.

On the left we have text to video examples, And on the right, we show the image to video examples.

And using SANA video as a base model, it can be applied to tasks like physical AI and real -world models.

Actually, by fine -tuning SANA video on a small data site with just a few safety steps,

we enable it to generate videos for robotic manipulation and gameplay play.

We also combine with the post -training tech in long lab, which is a three long test long method.

And this method is also accurate poster and is also from our group.

We also implement the DMD step distillation to further accelerate our model.

And finally, Sonar Video you can also generate minute -length video within just 36 seconds,

achieving 27 FPS real -time generation.

So, finally, there is one more thing.

Based on linear attention, we have much more potential applications to discover.

We believe the efficiency of linear attention

and deep compression auto encoder will facilitate streaming video generation

and also the development of long form like minute -length or even hour -length, real -world models.

And we are working on this project and we'll see if we can keep pushing the efficiency boundary.

And all of our projects, including the code and checkpoints,

100

are all open -source to encourage further exploration in efficient generative AI system.

101

So that's all for my talk today.

102

And we will have a poster session right after this aural, maybe in the afternoon.

103

Welcome everyone to discuss more.

104

Thanks so much.

105

Thank you.

106

Time for a quick question, if any.

107

Yeah, please go to the mic.

108

Thank you for your speech, and I have a question.

109

I think it's something similar to the auto -aggressive video diffusion models like cause of it

110

and something similar to the block causal linear attention you introduced.

111

So I wanted to ask what are the key differences and if we are thinking about infinite long video generation,

112

which one do you think would be better choice?

113

Thank you.

114

Thank you for your question.

115

Actually the cause V something like that, they actually perform on soft max extension.

116

So as I talked before, the soft max extension has an open square computation complexity.

117

So if you want to generate mini lens or even hour lens video, it's not possible to do it on soft max extension.

118

So if you want to generate mini lens video, so you have to keep the computation within like a sliding window or something like that.

119

So the main difference here is our linear tension has the global context.

120

It has memory from the first second to the final second.

121

So if you want to do some project or task like world model,

122

it has to be maintaining overall all the cache or the memory.

123

So I think that's the main advantages of getting attention.

124

Yeah, thank you.

125

Thank you.

تنزيل التطبيق

كل ما تحتاجه للتحدث بطلاقة

تقييم النطق بالذكاء الاصطناعيتقييم كل جملة

تدريب IPAأتقن كل صوت

المفرداتابنِ رصيدك من الكلمات

لعبة المفرداتتعلّم أثناء اللعب

امسح رمز QR

Download on theApp Store Get it onGoogle Play

FOR YOU

فيديو موصى به

747 جمل

MIT Introduction to Deep Learning | 6.S191

Daily ConversationUniversity Students

462 جمل

Expensive RTX 5090 for LLMs? NO. Use This Instead. (Tesla V100 SXM2 + Z8 G4, #RACERRRZ)

Daily ConversationUniversity StudentsWorking Adults

321 جمل

Mixture of Experts Explained Visually: How Trillion-Parameter Models Actually Work

Daily ConversationUniversity Students

28 جمل

Vanishing & Exploding Gradient explained | A problem resulting from backpropagation

Daily ConversationUniversity Students

75 جمل

What AI can (and can't) currently do in a real lab | Jennifer Doudna

Daily ConversationUniversity StudentsWorking Adults

المفردات وملاحظات التحدث لهذا الدرس

هذا الدرس في المحادثة مبني على الفيديو “SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer,”. أكثر الكلمات تكرارًا في الفيديو: block, attention, model, linear, generation. يحتوي هذا الفيديو على 125 جملة و1634 كلمة للتدرب على الترديد. مدة الكلام 12:46. يتكلم المتحدث بسرعة ثابتة، نحو 128 كلمة في الدقيقة، وهي سرعة مريحة للترديد. 80% فقط من الكلمات ضمن أكثر 3,000 كلمة شيوعًا في الإنجليزية، فالمفردات صعبة نسبيًا.

أهم المفردات في هذا الفيديو

15 كلمة من الفيديو تستحق التعلّم، مع النطق والمعنى:

الكلمة	النطق	المعنى
tension اسم	/ˈtɛnʃən/	تَوَتُّر
token اسم	/ˈtoʊkən/	رَمْز
efficient صفة	/ɪˈfɪʃənt/	فَعَّال
introduce فعل	/ˌɪntɹəˈdus/	قَدَّمَ
implement فعل	/ˈɪmpləmənt/	نَفَّذَ, نَجَزَ
lens اسم	/lɛnz/	عَدَسَة
rope اسم	/ɹoʊp/	حَبْل
contribution اسم	/ˌkɑntɹɪˈbjuʃən/	مُسَاهَمَة
explore فعل	/ɪkˈsploɹ/	اِسْتَكْشَفَ
poster اسم	/ˈpoʊstɚ/	مُلْصَق, أَفِيش
lean فعل	/ˈliːn/	مَالَ
combine فعل	/kəmˈbaɪn/	ضَمَّ, جَمَعَ
automobile اسم	/ˈɔː.tə.məˌbil/	سَيَّارَة, أُوتُومُوبِيل
module اسم	/ˈmɑd͡ʒul/	وَحْدَة
topic اسم	/ˈtɑpɪk/	مَوْضُوع, مَوَاضِيع

القواعد في هذا الفيديو

أكثر التراكيب التي يستخدمها المتحدث، مع العبارات نفسها من الفيديو:

التركيب	في الفيديو
المبني للمجهول be + اسم المفعول — التركيز على ما يحدث لا على من يفعله	are restricted · is based · are highlighted
جمل الوصل who / which + جملة — معلومة إضافية عن شخص أو شيء	two, which is · lab, which is

نطق يحتاج إلى انتباه

يستخدم المتحدث 4 من الصيغ المختصرة، مثل doesn't, I'm, we'll. انطقها بصيغتها القصيرة كما تسمعها.

أصوات “sh” و“zh”: cache /kæʃ/, tension /ˈtɛnʃən/, efficiency /ɪˈfɪʃ.ən.si/, extension /ɪkˈstɛnʃən/, efficient /ɪˈfɪʃənt/
الكلمات الطويلة — انتبه لموضع النبر: efficiency /ɪˈfɪʃ.ən.si/, contribution /ˌkɑntɹɪˈbjuʃən/, automobile /ˈɔː.tə.məˌbil/, potentially /pəˈtɛnʃ(ə)li/, mechanism /ˈmɛk.ə.nɪ.zəm/

أصوات يجدها متحدثو العربية صعبة:

/p/ — ليس /b/: انطقه مع نفخة هواء: temporal /ˈtɛm.pə.ɹəl/, input /ˈɪnpʊt/, implement /ˈɪmpləmənt/, rope /ɹoʊp/, explore /ɪkˈsploɹ/
/v/ — ليس /f/: مع اهتزاز الأوتار الصوتية: solve /sɒlv/, discover /dɪˈskʌvə/, aggressive /əˈɡɹɛs.ɪv/, leverage /ˈlɛv.(ə.)ɹɪd͡ʒ/, evolve /ɪˈvɑlv/
تتابع الصوامت — لا تضف حركة بينها: extension /ɪkˈstɛnʃən/, efficient /ɪˈfɪʃənt/, introduce /ˌɪntɹəˈdus/, implement /ˈɪmpləmənt/, lens /lɛnz/

كيف تتدرب بهذا الفيديو

استمع إلى الفيديو كاملًا مرة واحدة دون أن تتكلم، ودوّن الكلمات التي لا تعرفها.
ردّد جملة جملة بالسرعة العادية، وكرّر كل جملة حتى يتطابق إيقاعك مع المتحدث.
سجّل صوتك وقارنه بالأصل، مع الانتباه إلى كلمات مثل tension, token, efficient.

ما هي تقنية التظليل الصوتي؟

التظليل الصوتي (Shadowing) تقنية تعلم لغة مدعومة علمياً، طُورت أصلاً لتدريب المترجمين الفوريين المحترفين. الطريقة بسيطة لكنها قوية: تستمع لصوت إنجليزي أصلي وتكرره فوراً بصوت عالٍ — كظل يتبع المتحدث بتأخير 1-2 ثانية. تُظهر الأبحاث تحسناً كبيراً في دقة النطق والتنغيم والإيقاع وربط الأصوات والاستماع والطلاقة.

تقنية الشادوينغ: اقرأ الدليل الكامل خطوة بخطوة →

تدريب Shadowing: SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer, [ICLR 2026, Oral] - تعلم التحدث بالإنجليزية عبر الفيديو

اختر وضع التدريب

تنزيل التطبيق

MIT Introduction to Deep Learning | 6.S191

Expensive RTX 5090 for LLMs? NO. Use This Instead. (Tesla V100 SXM2 + Z8 G4, #RACERRRZ)

Mixture of Experts Explained Visually: How Trillion-Parameter Models Actually Work

Neural Radiance Fields in 2 Minutes

Controlled Multi-modal Image Generation for Plant Growth Modeling

Agentic AI Frameworks Explained: Workflows, Multi-Agent, & Production

Enhancing AV Simulation with Generative AI - DRIVE Labs Ep. 36

How I Learned Procedural Generation

The future unfolds! How to optimize Android apps for adaptive layouts

SAC | Soft Actor Critic (SAC) architecture | SAC Explained

Vanishing & Exploding Gradient explained | A problem resulting from backpropagation

What AI can (and can't) currently do in a real lab | Jennifer Doudna

المفردات وملاحظات التحدث لهذا الدرس

أهم المفردات في هذا الفيديو

القواعد في هذا الفيديو

نطق يحتاج إلى انتباه

كيف تتدرب بهذا الفيديو

ما هي تقنية التظليل الصوتي؟