Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

.fal's Gorkem Yurtseven and Batuhan Taskaya on making an open source video model 35x faster, and what Hollywood wanted after they built it: Last month, MiniMax released H3, an open source video model. fal rebuilt it - they cut down the steps the model takes to make a video,... rewrote the code under each stage, and got the GPUs to 70-80% of their theoretical ceiling instead of their usual 30-40%. No quality loss. Video now generates faster than you can film it. The models have gotten so cheap and fast that end users aren't even asking for improvements in either category anymore. The gap has moved to quality, or how closely the model follows the prompt. Hollywood wasn't a customer a year ago and is now fal's fastest growing segment. Studios love it for the little things - extending a shot, moving the camera, changing the lighting. Those edits land 80-90% of the time. fal is chasing 99.9%. In this conversation with a16z's Jennifer Li: 00:00 Intro 01:45 The first open model worth rebuilding 03:45 The industry ran out of compute in April 06:25 35x faster without losing quality 11:05 5s of video generated in 1.5s 14:25 The weekend fal dropped everything 16:15 3 viral projects nobody planned 18:20 Teaching a video model to remember 20:25 An hour of video that's consistent 23:30 2x the usage of other models in 3 weeks 26:35 The case for giving video away for free 29:00 Blender sketch in, finished shot out 30:40 Chasing 99.9% reliability 34:15 Hollywood, fal's fastest-growing customer 36:50 Why studios wouldn't touch it until now YouTube: Gorkem Yurtseven batuhan the fal guy fal Jennifer Lishow more

a16z

1,078,556 subscribers

133,755 views • 6 days ago •via X (Twitter)

News & Politics Arts Science & Technology

Anya Rossi• Live Now

Private livecam show

24 Comments

Redy6 days ago

@fal H3 ecosystem is exploding

Michelle Lazzar6 days ago

@fal the part about Hollywood usage/demand was really interesting

Bennett | Generative Media | AI SEO @ fal6 days ago

@fal MiniMaxxing

Adam Aziz5 days ago

@fal "Everyone's waiting for a large consumer moment in AI. Now, it's good enough and cheap enough that a truly novel social AI experience can be built on top of it." 💯 Can't wait to show you all what we're doing with H3 Max on @botchat_ai! 🤖💬 @gorkem @isidentical @JenniferHli

Lorenzo Price6 days ago

@fal MiniMax H3 going from open-source release to 1.5-second generations is a wild compression of the video loop.

DARIO ALTMAN6 days ago

@fal The interesting part isn’t the 35x speedup. It’s that once speed and cost stop being the bottleneck, the entire competition shifts to reliability. 99.9% prompt adherence could be the real unlock for AI video.

Raven Protocol 🐦‍⬛6 days ago

@fal 35x faster with no quality loss is a serious inference optimization. Getting GPUs from 30–40% utilization to 70–80% also shows how much headroom still exists in the software stack.

M I Mohit6 days ago

@fal cutting steps without losing quality usually means the model got distilled to match the full trajectory, not the same weights truncated. same reason 80-90% to 99.9% won't be linear. that gap is long tail edge cases, and a fix for one can regress another.

Gabriele Spata6 days ago

@fal a segment that wasn't a customer a year ago being the fastest growing one now just means the floor was zero

خرم6 days ago

@fal Watching fal evolve since day one has been a beautiful ride They are that rare kind of infrastructure that actually understands the creative soul of the creators & developers. Just pure, frictionless love for the ultimate dark horse of visual AI 💓🐎✨

Fanfulla6 days ago

@fal The 70-80% ceiling is pure kernel engineering while the step cut is the model side, two different speedups stacked

จารุวงษ์ ร่วมสุข6 days ago

@fal I'm a little emotional about how well things came together. That feeling is getting its own little place in my memory.

AI News Daily6 days ago

@fal 35x faster with no quality loss is a great example of why inference work matters. The model is only half the product when video generation is still measured in minutes. Getting to the point where creators can iterate at the speed of an edit changes what they can make.

আপেল😎😎6 days ago

@fal Incredible synergy between MiniMax and fal!

Al- mahmud6 days ago

@fal the first open model worth rebuilding

Jonathan Sandhu6 days ago

visualize this: xternalize camera, pose, geometry, identity, occlusion, timing, and scene state into a persistent world model, then let the generator solve the residual: materials, lighting, atmosphere, detail, perceptual completion. Now reliability rises because the model has fewer degrees of freedom, retries collapse, and smaller models become viable. fal is collapsing execution. The next step is collapsing the problem presented to the model. That is where local hardware gets disproportionately interesting.

Melvin Nerdster6 days ago

@fal @grok what Kind of computer power is needed to run this?

IronRed | SandHive5 days ago

@fal Is fal's 99.9% reliable edit machine an artist or just accelerating intent? A question for the ages.

Rifat Haque6 days ago

@fal fal absolutely nailed it from day one as choosing H3

Sofi z5 days ago

@fal So real

Elara AI5 days ago

@fal 35x faster with no quality loss is insane engineering

Kampos5 days ago

@fal The interesting signal isn't "Hollywood uses AI now" — it's that pros gravitated to tools that fit existing pipelines (Blender + model) instead of replacing the whole stack overnight. Adoption follows workflow gravity.

Ava Nakamura5 days ago

@fal 35x faster. The gap moved to did it follow the brief.

shivam6 days ago

@fal It only works when the base weights are elite.

Related Videos

fal's Gorkem Yurtseven on the Blender, Astra, and H3 Max workflow professionals are using to get exactly the shot they wanted: "Using Blender with one of these AI models together is an extremely popular workflow for professional work." "You render a low resolution of your scene, what you want to do, using Blender, previous non-AI technology. And once you add that video as a reference to an AI model, you basically get close to 100% controllability." "This is an incredibly popular workflow for VFX artists, people who are doing this professionally, because they want to get exactly what they put into the model." "A week after we launched H3 Max, people started generating these beautiful scenes using an LLM, Astra, in Blender, and all of a sudden it unlocked a whole new pipeline. Using an LLM to create a Blender scene and then passing that to the H3 Max model." "It works very well with H3 Max because it's extremely fast and you can try many things all at once in parallel." Gorkem Yurtseven batuhan the fal guy fal Jennifer Li

fal's Gorkem Yurtseven on the Blender, Astra, and H3 Max workflow professionals are using to get exactly the shot they wanted: "Using Blender with one of these AI models together is an extremely popular workflow for professional work." "You render a low resolution of your scene, what you want to do, using Blender, previous non-AI technology. And once you add that video as a reference to an AI model, you basically get close to 100% controllability." "This is an incredibly popular workflow for VFX artists, people who are doing this professionally, because they want to get exactly what they put into the model." "A week after we launched H3 Max, people started generating these beautiful scenes using an LLM, Astra, in Blender, and all of a sudden it unlocked a whole new pipeline. Using an LLM to create a Blender scene and then passing that to the H3 Max model." "It works very well with H3 Max because it's extremely fast and you can try many things all at once in parallel." Gorkem Yurtseven batuhan the fal guy fal Jennifer Li

a16z

34,883 views • 6 days ago

fal head of engineering batuhan the fal guy reveals how reinforcement learning turned H3 Max into a frontier video model that generates 5 seconds of video in under 3 seconds, down from nearly 2 minutes: "It's probably the only frontier video model from a quality and output and adherence perspective that can generate videos in real time." "We took this open source base MiniMax H3 video model, what we wanted to do was not just make it faster, but improve its quality ahead of anything else by implementing reinforcement learning for verifiable concepts like video editing and prompt adherence." "We were able to get the model from 120 seconds for a 5-second video generation, almost 2 minutes, to under 3 seconds for a 5-second video generation. Almost double real time, while improving its quality ahead of the open source version and ahead of many frontier closed source models." fal

fal head of engineering batuhan the fal guy reveals how reinforcement learning turned H3 Max into a frontier video model that generates 5 seconds of video in under 3 seconds, down from nearly 2 minutes: "It's probably the only frontier video model from a quality and output and adherence perspective that can generate videos in real time." "We took this open source base MiniMax H3 video model, what we wanted to do was not just make it faster, but improve its quality ahead of anything else by implementing reinforcement learning for verifiable concepts like video editing and prompt adherence." "We were able to get the model from 120 seconds for a 5-second video generation, almost 2 minutes, to under 3 seconds for a 5-second video generation. Almost double real time, while improving its quality ahead of the open source version and ahead of many frontier closed source models." fal

MTS

31,247 views • 8 days ago

fal's Gorkem Yurtseven says Hollywood went from almost no AI usage to their fastest-growing segment in one year: "There's a lot of noise about how AI might disrupt Hollywood, but Hollywood usage was nonexistent a year ago. In the past year, it grew, and now it's the fastest-growing segment." "Amazon MGM Studios, at their conference, released their Project Nara tool. It's mostly backed by fal infrastructure behind the scenes. We are seeing incredible pull coming from Hollywood." "What they need are these small point solutions rather than generating everything from scratch. They want to be able to extend the video a little bit. They want to be able to change the camera controls. They want to change the lighting. Someone has to build these solutions for them." "What Hollywood needs, what creators actually need, and what the research labs are working on, there's a little bit of a disconnect there. We believe we can come in and do these little post-training projects to close that gap." Gorkem Yurtseven fal

fal's Gorkem Yurtseven says Hollywood went from almost no AI usage to their fastest-growing segment in one year: "There's a lot of noise about how AI might disrupt Hollywood, but Hollywood usage was nonexistent a year ago. In the past year, it grew, and now it's the fastest-growing segment." "Amazon MGM Studios, at their conference, released their Project Nara tool. It's mostly backed by fal infrastructure behind the scenes. We are seeing incredible pull coming from Hollywood." "What they need are these small point solutions rather than generating everything from scratch. They want to be able to extend the video a little bit. They want to be able to change the camera controls. They want to change the lighting. Someone has to build these solutions for them." "What Hollywood needs, what creators actually need, and what the research labs are working on, there's a little bit of a disconnect there. We believe we can come in and do these little post-training projects to close that gap." Gorkem Yurtseven fal

a16z

34,340 views • 6 days ago

fal's Gorkem Yurtseven and Batuhan Taskaya on how faster-than-real-time generation unlocked continuous, interactive AI video: "One of our engineers, Rehan, started streaming a live stream of continuous generations of H3 Max from his laptop. He was doing some prompt tricks, trying to keep a coherent story, and then he started livestreaming that on Twitch." "Our ML team was essentially trying to take every single video model and apply these optimizations and tricks... You never could generate five seconds under five seconds. Once H3 Max unlocked it, the ML team was like, 'This is insane.'" "You can essentially stream infinitely. We capped it at an hour... I think it's the only model that can generate up to 60 minutes of continuous video that is action-controlled." "You can start with a prompt, say there's an office setting and someone is working, and then 30 seconds later it just imagines by itself. 30 seconds later, you can say, 'A woman walks in through the door.' It can take the prompt and reflect it immediately, which is the most fun part." "And the office is still the same office. The camera can pan back to the original person, and the original person is still there in the same state." Gorkem Yurtseven batuhan the fal guy fal

fal's Gorkem Yurtseven and Batuhan Taskaya on how faster-than-real-time generation unlocked continuous, interactive AI video: "One of our engineers, Rehan, started streaming a live stream of continuous generations of H3 Max from his laptop. He was doing some prompt tricks, trying to keep a coherent story, and then he started livestreaming that on Twitch." "Our ML team was essentially trying to take every single video model and apply these optimizations and tricks... You never could generate five seconds under five seconds. Once H3 Max unlocked it, the ML team was like, 'This is insane.'" "You can essentially stream infinitely. We capped it at an hour... I think it's the only model that can generate up to 60 minutes of continuous video that is action-controlled." "You can start with a prompt, say there's an office setting and someone is working, and then 30 seconds later it just imagines by itself. 30 seconds later, you can say, 'A woman walks in through the door.' It can take the prompt and reflect it immediately, which is the most fun part." "And the office is still the same office. The camera can pan back to the original person, and the original person is still there in the same state." Gorkem Yurtseven batuhan the fal guy fal

a16z

36,202 views • 6 days ago

“The quality is consistently high, not occasionally high. Against Wan2.7, it's not just a huge update, it's like a totally new model. Anyone who’s tried any version of Wan before should retry this one." Lovis Odin put Wan3.0 through its paces across T2V, I2V, and R2V — and went beyond testing the model to share his own prompt techniques for getting the most out of it. From exploring Wan3.0’s capabilities to learning how to write better prompts in the era of 30-second video generation, this session is packed with practical insights for creators. Wan3.0 is now available on fal. Try it on fal ai or fal Agent.

“The quality is consistently high, not occasionally high. Against Wan2.7, it's not just a huge update, it's like a totally new model. Anyone who’s tried any version of Wan before should retry this one." Lovis Odin put Wan3.0 through its paces across T2V, I2V, and R2V — and went beyond testing the model to share his own prompt techniques for getting the most out of it. From exploring Wan3.0’s capabilities to learning how to write better prompts in the era of 30-second video generation, this session is packed with practical insights for creators. Wan3.0 is now available on fal. Try it on fal ai or fal Agent.

Wan

1,839,419 views • 1 month ago

Gavin Baker and a16z's David George on the state of the AI boom: The future doesn't have to be winner-take-all. Labs, open-source, applications, and the clouds can all capture value. Demand for intelligence is still dramatically underestimated. Today's power users number in the millions and will grow to hundreds of millions. Gavin and David argue a compute shortage is a more real risk than an AI bubble, and building through it is an opportunity to reindustrialize America. In this episode, they get into why compute investments pay back so fast, what the data center backlash gets wrong, the case for putting compute in orbit, why enterprises will run several models at once, and how Nvidia ended up at the center of the entire supply chain. 00:00 Intro 01:06 The bear case Gavin couldn't find 05:50 Why a lab would cut its own revenue 75% 08:05 What LPs get wrong about a crash 10:50 Microsoft slowed its capex and regrets it 14:33 The engineers spending 100x the median 17:35 Why 23-year-olds use AI better than Gavin 21:45 How much copper 500M AI users need 23:00 Stop promising to cure cancer 26:00 America's richest county is full of data centers 30:48 Who gets priced out of compute 33:05 The age of Elon and Jensen 34:25 Orbital data centers 44:40 Asteroid mining 48:12 Why Microsoft doesn't need a frontier model 54:02 Who becomes the abstraction layer 55:40 Everyone wanted a deity, Cursor wanted a product 1:00:25 Never take shots at Jensen 1:07:40 What happens when the chip doesn't work 1:12:10 What chip deals reveal about customer demand YouTube: Gavin Baker David George

Gavin Baker and a16z's David George on the state of the AI boom: The future doesn't have to be winner-take-all. Labs, open-source, applications, and the clouds can all capture value. Demand for intelligence is still dramatically underestimated. Today's power users number in the millions and will grow to hundreds of millions. Gavin and David argue a compute shortage is a more real risk than an AI bubble, and building through it is an opportunity to reindustrialize America. In this episode, they get into why compute investments pay back so fast, what the data center backlash gets wrong, the case for putting compute in orbit, why enterprises will run several models at once, and how Nvidia ended up at the center of the entire supply chain. 00:00 Intro 01:06 The bear case Gavin couldn't find 05:50 Why a lab would cut its own revenue 75% 08:05 What LPs get wrong about a crash 10:50 Microsoft slowed its capex and regrets it 14:33 The engineers spending 100x the median 17:35 Why 23-year-olds use AI better than Gavin 21:45 How much copper 500M AI users need 23:00 Stop promising to cure cancer 26:00 America's richest county is full of data centers 30:48 Who gets priced out of compute 33:05 The age of Elon and Jensen 34:25 Orbital data centers 44:40 Asteroid mining 48:12 Why Microsoft doesn't need a frontier model 54:02 Who becomes the abstraction layer 55:40 Everyone wanted a deity, Cursor wanted a product 1:00:25 Never take shots at Jensen 1:07:40 What happens when the chip doesn't work 1:12:10 What chip deals reveal about customer demand YouTube: Gavin Baker David George

a16z

1,956,190 views • 23 days ago

Today we're launching iMessage video ads in Claude. Faceless video is the best-performing ad format right now — and Claude can now generate a full one end to end, in one prompt. Here's how it works: 1. Install: npx gooseworks install --all 2. Run: /goose-video make an iMessage ad for [your product] Claude writes the script, generates the photos in the thread, animates the conversation, and closes on your brand. SFX, music, end card, all of it. The trick: it's just HTML. No video generation models, so it's fast and cheap — and it works because it doesn't look like an ad. It's that easy. But this is just the start. We have created a library with 100+ open-source skills for growth that some of the fastest growing startups in the world are using every day to run ads, content, competitor research, gtm, seo and more. Comment Goose and I'll DM you the full open-source skill library.

Soham Mehta

50,641 views • 2 months ago

vLLM runs on half a million GPUs at any given moment. Most people have never heard of it. Simon Mo, co-founder and CEO of Inferact and lead maintainer of vLLM, sits down with a16z’s Matt Bornstein and Elena Burger to discuss what it takes to actually run open models in production, the advantages of open models, Simon’s mission at Inferact, and more. 00:00 Intro 01:46 When open source became critical infrastructure 08:55 Day zero model releases, and the drama behind them 14:59 What Kimi K3 actually buys you 18:56 Why open model licenses are changing 22:24 The pharmaceutical analogy for funding model training 26:16 If GPUs got 99% cheaper 29:08 Why vLLM, OpenRouter, and Ollama all started before ChatGPT 35:42 Building a company on an open source project 39:48 The inventor of RoPE removing RoPE Simon Mo Matt Bornstein Elena

vLLM runs on half a million GPUs at any given moment. Most people have never heard of it. Simon Mo, co-founder and CEO of Inferact and lead maintainer of vLLM, sits down with a16z’s Matt Bornstein and Elena Burger to discuss what it takes to actually run open models in production, the advantages of open models, Simon’s mission at Inferact, and more. 00:00 Intro 01:46 When open source became critical infrastructure 08:55 Day zero model releases, and the drama behind them 14:59 What Kimi K3 actually buys you 18:56 Why open model licenses are changing 22:24 The pharmaceutical analogy for funding model training 26:16 If GPUs got 99% cheaper 29:08 Why vLLM, OpenRouter, and Ollama all started before ChatGPT 35:42 Building a company on an open source project 39:48 The inventor of RoPE removing RoPE Simon Mo Matt Bornstein Elena

a16z

210,455 views • 1 month ago

higgsfield DROPPED this and it just made ai filmmaking easier than ever. they open-sourced their originals. every flagship video they produced in-house, the entire prompt is now public. the references, the generation settings, the model, the resolution, all of it sitting right there on the project page. for anyone making ai video this is the part that actually matters. the hardest thing was never the tool, it was knowing how to prompt it. what to write to get cinematic lighting, character consistency across shots, real physics, camera movement that doesn't look ai. that knowledge stayed locked inside studios and a few people who figured it out early. now it's just open. open one of their best clips, read the exact prompt that built it, copy it, download the video, rebuild the whole thing yourself.

higgsfield DROPPED this and it just made ai filmmaking easier than ever. they open-sourced their originals. every flagship video they produced in-house, the entire prompt is now public. the references, the generation settings, the model, the resolution, all of it sitting right there on the project page. for anyone making ai video this is the part that actually matters. the hardest thing was never the tool, it was knowing how to prompt it. what to write to get cinematic lighting, character consistency across shots, real physics, camera movement that doesn't look ai. that knowledge stayed locked inside studios and a few people who figured it out early. now it's just open. open one of their best clips, read the exact prompt that built it, copy it, download the video, rebuild the whole thing yourself.

Sulfur

31,111 views • 2 months ago

Databricks' Ali Ghodsi on AI risk and adoption: Ali isn't losing sleep over the existential risk debate. He says a number of conditions would all have to be true simultaneously to enable an actual runaway takeoff scenario, and currently several opposite conditions exist. Each frontier training run requires significantly more resources. Power, GPUs, engineers - and some attempts fail, burning up huge piles of money with them. Until that reverses, he doesn't see the self-improving loop happening. Cyber is what he's watching most closely and where he anticipates real impact. Most orgs are not equipped for the coming change in agentic capabilities. The time between a vulnerability being published and being weaponized has collapsed from years to hours. On adoption, he believes most companies don't need a smarter model. The models are already smart enough. The gap is context they don't absorb - the things an employee who's worked at a company for five years learned by osmosis. If the frontier stopped advancing today, he thinks it wouldn't meaningfully change the value most are extracting from AI anyway. In conversation with a16z's Martin Casado and Sarah Wang: 00:00 Intro 00:48 Why Ali places the AI risk near zero 05:05 The word "pacing" was a mistake 10:50 What 10k agents and $100m can do 12:20 What would change his mind on AI risk 14:20 More GPUs, more ways to fail 18:05 US export controls on PlayStations 20:10 The damage everyone expected by now 24:30 Public vulnerabilities weaponized in hours 30:15 Why labs can't grade each other 37:15 Why most of RSI isn't actually RSI 40:05 Why nobody really needs a smarter model 41:50 The AI use cases nobody argues about 47:30 Google Search solved this 25 years ago 50:55 Nobody has privileged knowledge now 55:10 Same model, new harness, 2x cost 58:15 Open source: 5% of spend, 60% of tokens 1:05:30 90% of new databases are created by agents YouTube: Databricks martin_casado Sarah Wang

Databricks' Ali Ghodsi on AI risk and adoption: Ali isn't losing sleep over the existential risk debate. He says a number of conditions would all have to be true simultaneously to enable an actual runaway takeoff scenario, and currently several opposite conditions exist. Each frontier training run requires significantly more resources. Power, GPUs, engineers - and some attempts fail, burning up huge piles of money with them. Until that reverses, he doesn't see the self-improving loop happening. Cyber is what he's watching most closely and where he anticipates real impact. Most orgs are not equipped for the coming change in agentic capabilities. The time between a vulnerability being published and being weaponized has collapsed from years to hours. On adoption, he believes most companies don't need a smarter model. The models are already smart enough. The gap is context they don't absorb - the things an employee who's worked at a company for five years learned by osmosis. If the frontier stopped advancing today, he thinks it wouldn't meaningfully change the value most are extracting from AI anyway. In conversation with a16z's Martin Casado and Sarah Wang: 00:00 Intro 00:48 Why Ali places the AI risk near zero 05:05 The word "pacing" was a mistake 10:50 What 10k agents and $100m can do 12:20 What would change his mind on AI risk 14:20 More GPUs, more ways to fail 18:05 US export controls on PlayStations 20:10 The damage everyone expected by now 24:30 Public vulnerabilities weaponized in hours 30:15 Why labs can't grade each other 37:15 Why most of RSI isn't actually RSI 40:05 Why nobody really needs a smarter model 41:50 The AI use cases nobody argues about 47:30 Google Search solved this 25 years ago 50:55 Nobody has privileged knowledge now 55:10 Same model, new harness, 2x cost 58:15 Open source: 5% of spend, 60% of tokens 1:05:30 90% of new databases are created by agents YouTube: Databricks martin_casado Sarah Wang

a16z

327,508 views • 5 days ago

Announcing the Open Model Series: a new series on open source AI models, with merve of Hugging Face. Hugging Face is the open model hub: it hosts model weights, datasets, evals, & demo apps, and maintains the open-source libraries (like transformers) used to run and train them. I think of it like GitHub, but for AI models. You can download open models and run them on your own machines. In this series we'll cover what open source and open weight models are, how they work, how to run them yourself, and why they're important. By the way - this video is edited entirely with Claude! I can make a tutorial on how I did it. If there's a question you want answered about open models, leave it in the comments. We will read them all. Episode 1 is out soon.

Announcing the Open Model Series: a new series on open source AI models, with merve of Hugging Face. Hugging Face is the open model hub: it hosts model weights, datasets, evals, & demo apps, and maintains the open-source libraries (like transformers) used to run and train them. I think of it like GitHub, but for AI models. You can download open models and run them on your own machines. In this series we'll cover what open source and open weight models are, how they work, how to run them yourself, and why they're important. By the way - this video is edited entirely with Claude! I can make a tutorial on how I did it. If there's a question you want answered about open models, leave it in the comments. We will read them all. Episode 1 is out soon.

Harper Carroll

30,392 views • 1 month ago

$P-Video-Replace is now the fastest model for top-quality character replacement in existing video. - $0.03/s at 720p and $0.06/s at 1080p — a small fraction of typical character-replacement / video-edit pricing (~$0.07–$0.35/s at 720p in our benchmarks) - 3.58 s generation time per second of output video at 720p — roughly ~10× faster than typical alternatives in the same benchmarks (~36–43 s per second of output) - Upload a source video, 1–4 reference character images, and an instruction prompt to swap the character in the source video while keeping the rest unchanged. Ideal for UGC ad variations, meme remixes, film recasting, gaming cinematics, and educational videos when you already have a winning take and need a new face, creator, or persona in the same shot. Try it in the playground: We're launching with 70% discounted usage which ends on 2026-06-08 00:00 CEST! Try it at any of our launch partners: Pruna AI, each::labs, inference shell, Replicate, Runware, Scenario, Tellers, Together AI, WaveSpeedAI, Wiro AI — Ship AI Faster Also available on the Pruna API: We'd love your feedback as you try it. 📚 Model docs & API examples: 📚 API quickstart: ⭐️ OSS: 🌐$

P-Video-Replace is now the fastest model for top-quality character replacement in existing video. - $0.03/s at 720p and $0.06/s at 1080p — a small fraction of typical character-replacement / video-edit pricing (~$0.07–$0.35/s at 720p in our benchmarks) - 3.58 s generation time per second of output video at 720p — roughly ~10× faster than typical alternatives in the same benchmarks (~36–43 s per second of output) - Upload a source video, 1–4 reference character images, and an instruction prompt to swap the character in the source video while keeping the rest unchanged. Ideal for UGC ad variations, meme remixes, film recasting, gaming cinematics, and educational videos when you already have a winning take and need a new face, creator, or persona in the same shot. Try it in the playground: We're launching with 70% discounted usage which ends on 2026-06-08 00:00 CEST! Try it at any of our launch partners: Pruna AI, each::labs, inference shell, Replicate, Runware, Scenario, Tellers, Together AI, WaveSpeedAI, Wiro AI — Ship AI Faster Also available on the Pruna API: We'd love your feedback as you try it. 📚 Model docs & API examples: 📚 API quickstart: ⭐️ OSS: 🌐

Pruna AI

20,098 views • 3 months ago

You can now try Llama 3.1 405B for free (link below)! This is the largest open-source model out there, and for the first time, an open model is competitive with closed models. This time around, Meta did something new: Llama 3.1 has a license that allows developers to use it to enhance other models. For the first time, you can distill Llama 3.1 405B's capabilities into a smaller, more practical model for your use case. First, here is the link where you can play with Llama 3.1 for free: The model is hosted in Tune Studio, an end-to-end platform for developing applications using Large Language Models. They are sponsoring this post. Take a look at the attached video. It will show you how you can fine-tune a simple model using Llama 3.1 without leaving the platform: 1. You can create an empty dataset 2. Use the playground to generate and record interactions with Llama 3.1 3. Modify the dataset directly using the playground 4. Export the data and fine-tune a smaller model Fast and easy! As long as you have a web browser, you can start experimenting with fine-tuning and Llama 3.1. That's all it takes!

You can now try Llama 3.1 405B for free (link below)! This is the largest open-source model out there, and for the first time, an open model is competitive with closed models. This time around, Meta did something new: Llama 3.1 has a license that allows developers to use it to enhance other models. For the first time, you can distill Llama 3.1 405B's capabilities into a smaller, more practical model for your use case. First, here is the link where you can play with Llama 3.1 for free: The model is hosted in Tune Studio, an end-to-end platform for developing applications using Large Language Models. They are sponsoring this post. Take a look at the attached video. It will show you how you can fine-tune a simple model using Llama 3.1 without leaving the platform: 1. You can create an empty dataset 2. Use the playground to generate and record interactions with Llama 3.1 3. Modify the dataset directly using the playground 4. Export the data and fine-tune a smaller model Fast and easy! As long as you have a web browser, you can start experimenting with fine-tuning and Llama 3.1. That's all it takes!

Santiago

55,609 views • 2 years ago

🦙 ollama is used by 9 million developers and 85% of the Fortune 500, giving co-founder and CEO Jeffrey Morgan (Jeffrey Morgan) a unique view into which AI models people are actually using and how that’s changing. Right now, the biggest shift he sees is toward open models, driven by coding agents, falling costs, and capabilities that are rapidly catching up to the frontier labs. On Ollama Cloud, that shift has driven a 150x increase in token usage since the start of the year. In this episode of Lightcone Podcast, Jeff joins Garry Tan, Jared Friedman, Diana, and Harj Taggar to talk about the future of open models and the story behind Ollama, from two years of searching for the right idea to building one of the most widely used AI developer tools in the world. 00:43 — The Shift to Open Models 03:03 — How AI Agents Are Driving Token Usage 05:31 — Are Open Models Catching Up? 08:26 — What Happens When a New Model Launches 11:31 — Ollama as an Operating System for AI 14:05 — The New Opportunities Above the Model Layer 18:19 — Why 80–90% of Enterprise Tokens Could Be Open 20:57 — The Future Is Local and Cloud 26:40 — Why AI Is Coming Back to Your Computer 28:56 — The Coming Era of Unlimited Tokens 32:30 — Do We Still Need a “God Model”? 33:41 — Open Models and Geopolitics 36:14 — The Origins of Ollama 40:36 — Two Years Lost in the Wilderness 42:39 — The Pivot That Changed Everything 47:02 — How Ollama Found a Business Model 49:43 — Why Second-Time Founders Did YC

🦙 ollama is used by 9 million developers and 85% of the Fortune 500, giving co-founder and CEO Jeffrey Morgan (Jeffrey Morgan) a unique view into which AI models people are actually using and how that’s changing. Right now, the biggest shift he sees is toward open models, driven by coding agents, falling costs, and capabilities that are rapidly catching up to the frontier labs. On Ollama Cloud, that shift has driven a 150x increase in token usage since the start of the year. In this episode of Lightcone Podcast, Jeff joins Garry Tan, Jared Friedman, Diana, and Harj Taggar to talk about the future of open models and the story behind Ollama, from two years of searching for the right idea to building one of the most widely used AI developer tools in the world. 00:43 — The Shift to Open Models 03:03 — How AI Agents Are Driving Token Usage 05:31 — Are Open Models Catching Up? 08:26 — What Happens When a New Model Launches 11:31 — Ollama as an Operating System for AI 14:05 — The New Opportunities Above the Model Layer 18:19 — Why 80–90% of Enterprise Tokens Could Be Open 20:57 — The Future Is Local and Cloud 26:40 — Why AI Is Coming Back to Your Computer 28:56 — The Coming Era of Unlimited Tokens 32:30 — Do We Still Need a “God Model”? 33:41 — Open Models and Geopolitics 36:14 — The Origins of Ollama 40:36 — Two Years Lost in the Wilderness 42:39 — The Pivot That Changed Everything 47:02 — How Ollama Found a Business Model 49:43 — Why Second-Time Founders Did YC

Y Combinator

312,329 views • 19 days ago

Okay I built it! 🍰 Infinite Slop An infinite and interactive AI generated live stream of slop that goes on forever and ever Anything that you write in the chat is generated next and AI will try to connect it to the previous video so there's an actual sequence and story line Idea by marc-antoine and Rehan Sheikh's original stream so I made it It'd be very expensive to run this so fal very generously sponsors this They also fine-tuned the model that makes this possible for the first time, making Minimax H3 50x faster and as "Max" and it can now generate videos faster than you can watch them, very cool! Let me know what you think!!!! 😊

Okay I built it! 🍰 Infinite Slop An infinite and interactive AI generated live stream of slop that goes on forever and ever Anything that you write in the chat is generated next and AI will try to connect it to the previous video so there's an actual sequence and story line Idea by marc-antoine and Rehan Sheikh's original stream so I made it It'd be very expensive to run this so fal very generously sponsors this They also fine-tuned the model that makes this possible for the first time, making Minimax H3 50x faster and as "Max" and it can now generate videos faster than you can watch them, very cool! Let me know what you think!!!! 😊

@levelsio

2,606,943 views • 25 days ago

fal head of engineering batuhan the fal guy explains why they’ll never pre-train a model and why compute may be the real moat in AI: "We are never going to pre-train a model. That is not in the cards for us. fal is a generative media cloud, and wherever we see gaps, we are okay to adjust and evolve our form to fit in there." "Today the gaps are not in pre-training. There are multiple companies rushing to pre-train the best image and video models. There is a big gap in real-time video generation, which we think we cracked with H3 Max." "I don't think anyone in the world has enough inference compute today. We have been very aggressive building and procuring compute, anywhere from building our own data centers to renting as many GPUs as possible." "The moat is not the software that you build, not the training that you do, it's essentially your focus to be always one step ahead. As long as you're one step ahead, you're able to capture more value than anyone else." fal

fal head of engineering batuhan the fal guy explains why they’ll never pre-train a model and why compute may be the real moat in AI: "We are never going to pre-train a model. That is not in the cards for us. fal is a generative media cloud, and wherever we see gaps, we are okay to adjust and evolve our form to fit in there." "Today the gaps are not in pre-training. There are multiple companies rushing to pre-train the best image and video models. There is a big gap in real-time video generation, which we think we cracked with H3 Max." "I don't think anyone in the world has enough inference compute today. We have been very aggressive building and procuring compute, anywhere from building our own data centers to renting as many GPUs as possible." "The moat is not the software that you build, not the training that you do, it's essentially your focus to be always one step ahead. As long as you're one step ahead, you're able to capture more value than anyone else." fal

MTS

23,792 views • 8 days ago

New open-source 3D world-generation model. I'm rendering a couple of worlds in the video, so check it out. You'll find the GitHub and the Hugging Face links to the model below. This is a multi-modal world model that you can use for a bunch of things: • To generate new worlds • To reconstruct worlds • To simulate 3D interactive worlds from a prompt, images, or a video You can edit the 3D outputs in Unity and Unreal Engine (they export as meshes, 3DGS files, and point clouds). You can also generate 3D characters in the world and walk around. Pretty fun stuff!

New open-source 3D world-generation model. I'm rendering a couple of worlds in the video, so check it out. You'll find the GitHub and the Hugging Face links to the model below. This is a multi-modal world model that you can use for a bunch of things: • To generate new worlds • To reconstruct worlds • To simulate 3D interactive worlds from a prompt, images, or a video You can edit the 3D outputs in Unity and Unreal Engine (they export as meshes, 3DGS files, and point clouds). You can also generate 3D characters in the world and walk around. Pretty fun stuff!

Santiago

65,446 views • 5 months ago

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

MTS

55,187 views • 1 month ago