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.

🤩 75% cheaper Image upscaling 🤑 GPU costs for image upscaling from 1 to 4 Megapixels are less then a penny now 💵 Since the speed improvements I made, Clarity Upscaler now runs just as fast on both the cheaper A40 and the pricier A100 GPUs. The total GPU... show more

philz1337x

17,401 subscribers

69,210 views • 2 years ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

10 Comments

philz1337x2 years ago

Me in the past: 👨‍💻 Coding for a week and only telling another nerd what I did Me now: 🥳 Choosing another GPU on Replicate and making a huge announcement with video and 10 emojis 😄

Heba AI2 years ago

Left original, right from replicate with default settings. What parameter needs tweaking? Love how it changed the text in shirt from Gluteous Maximus to Sluteous Maximus 🤣

Floris2 years ago

Would it work on sequential video frames? Would love to see that!

philz1337x2 years ago

I will try

𝑨𝒓𝒕𝒊𝒇𝒊𝒄𝒊𝒂𝒍 𝑮𝒖𝒚2 years ago

You tried to run it with SD light/turbo?? Maybe can improve.. Or Maybe trying some different schedule + steps calp help improve the speed.

philz1337x2 years ago

Good idea, I didn't tried yet. As far as I know SD light and turbo is only available for SDXL and clarity uses SD1.5. I could try LCM but I think the quality is not 100% the same.

Su2 years ago

Mate, this is exactly what I have been waiting for! Thanks for the hard work!

Dawid Ryś2 years ago

Will it run locally on RTX cards like 4070?

🅻🅾🅵🅸9112 years ago

It doesnt just upscale. It very often completely changes the main parts of an image. How many gigs (vram) does it require when using locally ?

philz1337x2 years ago

Yes it’s also able to enhance the image. You can reduce the effect by lowering creativity. You can reduce VRAM usage with VAE tiling.

Related Videos

🤩 30% more speed and increased quality in high-resolution image upscaling I've improved clarity upscaling model by reducing the creativity by 20% in every internal upscaling step. It makes the amount of hallucination much more controllable, thus increasing the quality while also being 30% faster ⏩Now live: App: Code: Replicate: 🏞️ It's great how the blurry background stays blurry, but the cat becomes super sharp This example was made with a creativity setting of 0.25, and you can go lower if you want less hallucination

🤩 30% more speed and increased quality in high-resolution image upscaling I've improved clarity upscaling model by reducing the creativity by 20% in every internal upscaling step. It makes the amount of hallucination much more controllable, thus increasing the quality while also being 30% faster ⏩Now live: App: Code: Replicate: 🏞️ It's great how the blurry background stays blurry, but the cat becomes super sharp This example was made with a creativity setting of 0.25, and you can go lower if you want less hallucination

philz1337x

28,236 views • 2 years ago

Crystal Upscaler is now up to 10× faster 🤯 Upscaling to 1K took 12 s → now 1.2 s 🤩 2K: 16 s → 4 s 4K: 47 s → 15 s 5K: 71 s → 21 s I tracked down everything that slowed down the upscale, and now it’s finally using the full potential of the H100 GPU 🔥 Next step: increasing the 5K limit and maybe video upscaling?

Crystal Upscaler is now up to 10× faster 🤯 Upscaling to 1K took 12 s → now 1.2 s 🤩 2K: 16 s → 4 s 4K: 47 s → 15 s 5K: 71 s → 21 s I tracked down everything that slowed down the upscale, and now it’s finally using the full potential of the H100 GPU 🔥 Next step: increasing the 5K limit and maybe video upscaling?

philz1337x

299,157 views • 11 months ago

Crystal Upscaler now supports 10k 🤩 It works because I added the tiling system from Clarity Upscaler Now live on the Clarity AI API and dashboard

Crystal Upscaler now supports 10k 🤩 It works because I added the tiling system from Clarity Upscaler Now live on the Clarity AI API and dashboard

philz1337x

233,795 views • 11 months ago

✨ I increased the max resolution for my open-source upscaler to 20,000 x 20,000 pixels (or 400 megapixels) 🤯 This makes it by far the highest resolution AI upscaler available It takes 26 minutes to run on an A100 GPU 80GB and produces a 100MB JPG file 😱 The higher upscale resolution is possible because I am running a second replicate model now which runs on a larger GPU with 80GB VRAM I would like to increase the resolution even higher but it runs out of VRAM for higher resolutions which I might be able to fix by changing the tiling settings 🛠️ My code is open-source, and I also host a very easy-to-use version on my website Clarity AI, where users can upscale anything between 2 and 20,000 pixels now

✨ I increased the max resolution for my open-source upscaler to 20,000 x 20,000 pixels (or 400 megapixels) 🤯 This makes it by far the highest resolution AI upscaler available It takes 26 minutes to run on an A100 GPU 80GB and produces a 100MB JPG file 😱 The higher upscale resolution is possible because I am running a second replicate model now which runs on a larger GPU with 80GB VRAM I would like to increase the resolution even higher but it runs out of VRAM for higher resolutions which I might be able to fix by changing the tiling settings 🛠️ My code is open-source, and I also host a very easy-to-use version on my website Clarity AI, where users can upscale anything between 2 and 20,000 pixels now

philz1337x

165,069 views • 1 year ago

✨ Clarity Upscaler now has a Pro version It's the first upscaler that can add realism without changing a person's identity Two years ago I launched Clarity Upscaler, which has since made 30 million runs. I still can't really believe it 🤩 It's great for creative upscaling but it always changes people's identity too much. I made Crystal to solve that, but it's not able to add details or realism. Now we have Clarity Pro Upscaler which finally solves all that and can enhance images without changing the content. In the coming days I will make it even more powerful by adding parameters and increasing the max resolution even further (it's already at 8K!) Happy to hear your feedback

✨ Clarity Upscaler now has a Pro version It's the first upscaler that can add realism without changing a person's identity Two years ago I launched Clarity Upscaler, which has since made 30 million runs. I still can't really believe it 🤩 It's great for creative upscaling but it always changes people's identity too much. I made Crystal to solve that, but it's not able to add details or realism. Now we have Clarity Pro Upscaler which finally solves all that and can enhance images without changing the content. In the coming days I will make it even more powerful by adding parameters and increasing the max resolution even further (it's already at 8K!) Happy to hear your feedback

philz1337x

50,179 views • 5 months ago

Crystal is now the largest image-upscaler API in the world I've increased the output limit to 1.5 gigapixels, which is 38K 😂

Crystal is now the largest image-upscaler API in the world I've increased the output limit to 1.5 gigapixels, which is 38K 😂

philz1337x

156,803 views • 3 months ago

#MiniFlix App Features ⚡️ 💎We've made significant improvements to MiniFlix as a #ShortDrama App to watch, earn and enjoy a seamless #Web3 experience 🪂 😍Combining 👉Shortdrama, Web3, #AIChat, Copyright contents and more 🔥 📍Get the App NOW using the QR code on the website!

#MiniFlix App Features ⚡️ 💎We've made significant improvements to MiniFlix as a #ShortDrama App to watch, earn and enjoy a seamless #Web3 experience 🪂 😍Combining 👉Shortdrama, Web3, #AIChat, Copyright contents and more 🔥 📍Get the App NOW using the QR code on the website!

Mini Flix

26,693 views • 1 year ago

American companies have moved on from just shrinking their products Now they are shrinking their products AND decreasing the primary ingredients inside Example: Tyson chicken nuggets - The bag used to be 2 pounds, not it’s 29 ounces. Thats 11% less for the same price - The serving size used to be 270 calories, now for the same amount of nuggets you get 210 calories and less protein - How they did that was, they decreased the amount of chicken you get in each nugget and added more water and fiber So now Americans are getting robbed with both “Shrinkflation” and “Skimflation” We keep paying more for less

American companies have moved on from just shrinking their products Now they are shrinking their products AND decreasing the primary ingredients inside Example: Tyson chicken nuggets - The bag used to be 2 pounds, not it’s 29 ounces. Thats 11% less for the same price - The serving size used to be 270 calories, now for the same amount of nuggets you get 210 calories and less protein - How they did that was, they decreased the amount of chicken you get in each nugget and added more water and fiber So now Americans are getting robbed with both “Shrinkflation” and “Skimflation” We keep paying more for less

Wall Street Apes

71,741 views • 7 months ago

Clarity Pro Upscaler now supports up to 10K That's 100 Megapixels and a single PNG is already ~130 MB 😅

Clarity Pro Upscaler now supports up to 10K That's 100 Megapixels and a single PNG is already ~130 MB 😅

philz1337x

47,895 views • 5 months ago

$23 to upscale one image 😱 I have created the most expensive upscaler in the world 😅

$23 to upscale one image 😱 I have created the most expensive upscaler in the world 😅

philz1337x

523,628 views • 3 months ago

CoreWeave co-founder brannin mcbee reveals the A100 demand through 2029 is a big narrative violation for a market that thought GPUs would only have 2 to 3 years of useful life: "You don't need the latest generation GPU for every workload. You can think about it as a matrix of different sizes of models versus different sizes of GPUs, and it all just gets filled in volumetrically and economically." "That A100 GPU will cost less on a dollar per hour basis than a brand-new GB300 GPU. But the workload fits most efficiently on that A100, and it ultimately comes down to the engineering organizations making those requests." "It's been a pretty big narrative violation for the market that thought that these GPUs would only have two to three years of useful life, and instead we're contracting GPUs out through its ninth year of life. We expect to see that repeated across subsequent SKUs as well." CoreWeave

MTS

11,699 views • 1 month ago

THIS DEVELOPER JUST RAN A TRILLION PARAMETER MODEL ON 4 MAC STUDIOS - 10X FASTER AND 5X CHEAPER THAN CLOUD CODE 19:00 he says it out loud. "we just ran a trillion parameter model. 30 something tokens per second. wow." RDMA over Thunderbolt made the cluster 10x faster than before. tensor parallelism splits the model between machines in parallel instead of sequentially DeepSeek V3 full size without quantization. 183GB per machine. still 25 tokens per second. faster than most people can read Kimi K2 with trillion parameters and 256,000 context window. loaded, responded and adapted memory usage dynamically based on prompt size 4 machines consuming 66 watts total. cloud GPU for the same workload costs $1,900/month a cluster that used to cost millions now assembled from Mac Studios in a few hours

THIS DEVELOPER JUST RAN A TRILLION PARAMETER MODEL ON 4 MAC STUDIOS - 10X FASTER AND 5X CHEAPER THAN CLOUD CODE 19:00 he says it out loud. "we just ran a trillion parameter model. 30 something tokens per second. wow." RDMA over Thunderbolt made the cluster 10x faster than before. tensor parallelism splits the model between machines in parallel instead of sequentially DeepSeek V3 full size without quantization. 183GB per machine. still 25 tokens per second. faster than most people can read Kimi K2 with trillion parameters and 256,000 context window. loaded, responded and adapted memory usage dynamically based on prompt size 4 machines consuming 66 watts total. cloud GPU for the same workload costs $1,900/month a cluster that used to cost millions now assembled from Mac Studios in a few hours

Noisy

80,087 views • 3 months ago

PowerInfer - a high-speed inference engine for deploying LLMs locally. Just came across this super interesting project on speeding up inference. It's not MoE but it's a simple approach that exploits the high locality in LLM inference to design a GPU-CPU hybrid inference engine. Hot-activated neurons are preloaded onto the GPU for fast access, while cold-activated neurons (the majority) are computed on the CPU. This approach significantly reduces GPU memory demands and CPU-GPU data transfer. It achieves an average token generation rate of 13.20 tokens/s, with a peak of 29.08 tokens/s, across various LLMs on a single NVIDIA RTX 4090 GPU. It's on only 18% lower than that achieved by a top-tier server-grade A100 GPU. It also significantly outperforms llama.cpp by up to 11.69x while retaining model accuracy. There is a lot more innovation around inference that's coming fast. Really encouraged by the study on sparse computation to enhance the computational efficiency of LLMs. It's now possible to use PowerInfer with Llama 2 and Faclon 40B. Mistral-7B support is coming soon!

PowerInfer - a high-speed inference engine for deploying LLMs locally. Just came across this super interesting project on speeding up inference. It's not MoE but it's a simple approach that exploits the high locality in LLM inference to design a GPU-CPU hybrid inference engine. Hot-activated neurons are preloaded onto the GPU for fast access, while cold-activated neurons (the majority) are computed on the CPU. This approach significantly reduces GPU memory demands and CPU-GPU data transfer. It achieves an average token generation rate of 13.20 tokens/s, with a peak of 29.08 tokens/s, across various LLMs on a single NVIDIA RTX 4090 GPU. It's on only 18% lower than that achieved by a top-tier server-grade A100 GPU. It also significantly outperforms llama.cpp by up to 11.69x while retaining model accuracy. There is a lot more innovation around inference that's coming fast. Really encouraged by the study on sparse computation to enhance the computational efficiency of LLMs. It's now possible to use PowerInfer with Llama 2 and Faclon 40B. Mistral-7B support is coming soon!

elvis

261,756 views • 2 years ago

P-Image-Upscale is the fastest and cheapest image upscaler in the world: supporting outputs up to 128 MP under 1 seconds. A couple of weeks ago we released p-image-upscale, and it now works better than ever. It’s the fastest image upscaler in the world, supporting outputs up to 128 MP while keeping pricing simple and predictable: - $0.005/image for 1–4 MP - $0.01/image for 4–8 MP - $0.02/image for 8–16 MP - $0.04/image for 16–32 MP - $0.06/image for 32–64 MP - $0.12/image for 64–128 MP That means you can go from low-res input to production-ready output with extreme speed, while preserving detail and keeping costs easy to understand. Available on: - Pruna AI | - each::labs | - inference shell | - Replicate | - Runware | - Segmind | - WaveSpeedAI | - Wiro AI — Ship AI Faster | If you’re building workflows where image quality, price, latency, and scale all matter, p-image-upscale is built for you.

P-Image-Upscale is the fastest and cheapest image upscaler in the world: supporting outputs up to 128 MP under 1 seconds. A couple of weeks ago we released p-image-upscale, and it now works better than ever. It’s the fastest image upscaler in the world, supporting outputs up to 128 MP while keeping pricing simple and predictable: - $0.005/image for 1–4 MP - $0.01/image for 4–8 MP - $0.02/image for 8–16 MP - $0.04/image for 16–32 MP - $0.06/image for 32–64 MP - $0.12/image for 64–128 MP That means you can go from low-res input to production-ready output with extreme speed, while preserving detail and keeping costs easy to understand. Available on: - Pruna AI | - each::labs | - inference shell | - Replicate | - Runware | - Segmind | - WaveSpeedAI | - Wiro AI — Ship AI Faster | If you’re building workflows where image quality, price, latency, and scale all matter, p-image-upscale is built for you.

Pruna AI

24,089 views • 4 months ago

QVAC SDK 0.11.0 is live. 🛠️ This release focuses entirely on unlocking next-generation local compute and advanced visual workflows. What’s new: Next-Gen Models: Core engine updated to the latest version of Fabric, unlocking full support for Qwen 3.5, Qwen 3.6, and Gemma 4. Multi-GPU Support: The SDK can now split workloads across multiple graphics cards on the same machine, allowing you to run significantly larger models completely locally. Multi-Image Conditioning: Blend multiple reference images together in a single generation for advanced style mixing and composition control. On-Device Upscaling: Boost your generated images to high-quality resolutions, running securely on your own hardware. More improvements are waiting under the hood. Check the change logs, update your SDK today, and start building with

QVAC SDK 0.11.0 is live. 🛠️ This release focuses entirely on unlocking next-generation local compute and advanced visual workflows. What’s new: Next-Gen Models: Core engine updated to the latest version of Fabric, unlocking full support for Qwen 3.5, Qwen 3.6, and Gemma 4. Multi-GPU Support: The SDK can now split workloads across multiple graphics cards on the same machine, allowing you to run significantly larger models completely locally. Multi-Image Conditioning: Blend multiple reference images together in a single generation for advanced style mixing and composition control. On-Device Upscaling: Boost your generated images to high-quality resolutions, running securely on your own hardware. More improvements are waiting under the hood. Check the change logs, update your SDK today, and start building with

QVAC

2,006,449 views • 4 months ago

OPENAI JUST CUT MODEL PRICES AS THE AI PRICE WAR HEATS UP 🔥 OpenAI is cutting the price of Luna by 80% and Terra by 20%. It is also rolling out a faster mode that runs at twice the speed, but costs more. The bigger picture is clear -- cheaper Chinese models and tighter corporate AI budgets are forcing the biggest AI companies to compete on price, not just performance.

OPENAI JUST CUT MODEL PRICES AS THE AI PRICE WAR HEATS UP 🔥 OpenAI is cutting the price of Luna by 80% and Terra by 20%. It is also rolling out a faster mode that runs at twice the speed, but costs more. The bigger picture is clear -- cheaper Chinese models and tighter corporate AI budgets are forcing the biggest AI companies to compete on price, not just performance.

CryptosRus

12,382 views • 2 months ago

Image. Then motion. GPT Image 2 + Seedance 2.0 are now both live on Pollo AI. GPT Image 2 → structured layouts, sharp text, consistent characters — ready for ads, products, and social content. Seedance 2.0 → cinematic, stable, smooth video — straight from your image. From image to motion, the workflow is fast, seamless, and powerful.

Image. Then motion. GPT Image 2 + Seedance 2.0 are now both live on Pollo AI. GPT Image 2 → structured layouts, sharp text, consistent characters — ready for ads, products, and social content. Seedance 2.0 → cinematic, stable, smooth video — straight from your image. From image to motion, the workflow is fast, seamless, and powerful.

Pollo AI

432,024 views • 5 months ago

✨ For the last few months I have been reverse engineering Magnific AI's famous upscaler It uses MultiDiffusion, ControlNet tiles and details LoRas In true AI spirit, I am open sourcing it for everyone to use for free in your apps It's designed to easily control the amount of hallucination, resemblance and creativity. You can use it on Replicate now: Or run it locally with A1111'a web UI (parameters are in the comments) The code is fully open source:

✨ For the last few months I have been reverse engineering Magnific AI's famous upscaler It uses MultiDiffusion, ControlNet tiles and details LoRas In true AI spirit, I am open sourcing it for everyone to use for free in your apps It's designed to easily control the amount of hallucination, resemblance and creativity. You can use it on Replicate now: Or run it locally with A1111'a web UI (parameters are in the comments) The code is fully open source:

philz1337x

565,773 views • 2 years ago