The AI discussion thread

[Kaido]

Hunyuan Image 3.0 enters the image-generation fray!

GitHub - Tencent-Hunyuan/HunyuanImage-3.0: HunyuanImage-3.0: A Powerful Native Multimodal Model for Image Generation

HunyuanImage-3.0: A Powerful Native Multimodal Model for Image Generation - Tencent-Hunyuan/HunyuanImage-3.0

github.com

腾讯混元

腾讯研发的大语言模型

hunyuan.tencent.com

FREE & OPEN SOURCE!! Beating out Nano Banana!

Tencent’s AI image model beats Google’s Nano Banana in leaderboard

Released late in September, Hunyuan Image 3.0 is the industry’s biggest open-source image-generation model to date.

www.scmp.com

September 28, 2025 — Tencent HunYuan today announced and open-sourced HunYuanImage 3.0, a native multimodal image generation model with 80B parameters. HunYuanImage 3.0 is the first open-source, industrial-grade native multimodal text-to-image model and currently the best-performing and largest open-source image generator, benchmarking against leading closed-source systems.

Users can try HunYuanImage 3.0 on the desktop version of the Tencent HunYuan website Tensor.Art (https://tensor.art) is soon to support online generation! The model will also roll out on Yuanbao. Model weights and accelerated builds are available on GitHub and Hugging Face; both enterprises and individual developers may download and use them free of charge.

HunYuanImage 3.0 brings commonsense and knowledge-based reasoning, high-accuracy semantic understanding, and refined aesthetics that produce high-fidelity, photoreal images. It can parse thousand-character prompts and render long text inside images—delivering industry-leading generation quality.

More reading on the model:

https://www.reddit.com/r/TensorArt_HUB/comments/1nsihsd/the_best_performing_open_source_image_generation

 

[RnR_au]

Yeah the Chinese models you can run at home is capitalism's worst nightmare

edit:... just saw the memory requirements for this model "GPU Memory: ≥3×80GB (4×80GB recommended for better performance)"

 

[Kaido]

Midjourney for base images opens up so many doors for cool videos:

https://twitter.com/x/status/1976636672441032892

 

[Kaido]

Neat!

https://finance.yahoo.com/news/plumber-favorite-tool-chatgpt-090052048.html

A survey of more than 400 tradespeople across North America by Housecall Pro earlier this year found that more than 70% of respondents said they have tried AI tools and about 40% said they actively use them. Younger professionals are leading the charge, the survey showed, though older workers are testing the waters, too.

 

[Kaido]

Cyberpunk Halloween!

https://www.reddit.com/r/midjourney/comments/1n3obik/cyberpunk_market

 

[igor_kavinski]

Pretty awesome!

https://www.reddit.com/user/Endlesstavernstiktok/

Especially

https://www.reddit.com/r/midjourney/comments/1o2xn22/how_2_krump

Let's ROCK!

 

[Kaido]

Animation is getting REALY fun!

View this content on Instagram

View this content on Instagram

 

[igor_kavinski]

EPIC

https://www.reddit.com/r/dndai/comments/1nkpjc8/the_song_of_creation_bahamut_x_tiamat_duet

 

[Kaido]

Midjourney for base images opens up so many doors for cool videos:

https://twitter.com/x/status/1976636672441032892

 

[Kaido]

So here's an interesting look into the future: Real-time interactive simulation

1. Multi-modal input means we can use Video-to-Video style transfer, such as AI upscalng & generative quality improvement.
2. Real-time reskinning now exists, just not at a high-quality level. This can be applied to both existing video games AND live video!

Imagine real-time AI-scaling in the future over older & low-poly games:

https://twitter.com/x/status/1977339952460632145

Decart has the base real-time rskinning technology up & running with VR support with camera pass-thru:

https://twitter.com/x/status/1975219882280223031

Decart API Platform

Creativity without the wait

platform.decart.ai

 

[Kaido]

So here's an interesting look into the future: Real-time interactive simulation

1. Multi-modal input means we can use Video-to-Video style transfer, such as AI upscalng & generative quality improvement.
2. Real-time reskinning now exists, just not at a high-quality level. This can be applied to both existing video games AND live video!

Imagine real-time AI-scaling in the future over older & low-poly games:

Decart has the base real-time rskinning technology up & running with VR support with camera pass-thru:

Sooooo many applications:

1. Live video streaming, Zoom meetings, Facetime effects, Snapchat filters, etc.
2. Post-production video processinh
3. Older game uprezzing
4. Low-poly, low-GPU-demand upscaling & reskinning
5. VR games & video pass-thru enhancement

https://twitter.com/x/status/1973125823952777720

https://twitter.com/x/status/1973251062305005964

 

[Kaido]

One of my biggest interests with AI is in LIDAR, for 2 applications:

1. Self-driving cars
2. Ground-penetrating archeological mapping

Waymo & other companies are doing some neat stuff with LIDAR & AI. But what's even more fun is aerial LIDAR-mapping of hidden historical archeological structures, especially in South America. Great background story here:

Airborne Lidar for Archaeology in Central and South America - LIDAR Magazine

The rainforests in Central and South America give little indication of the civilizations that they’ve swallowed up. A few centuries ago these regions...

lidarmag.com

The first commercial lidar sensors became available in the mid-1990s. Unlike traditional photographic sensors, airborne lidar had the unique capability to be used day or night, penetrate vegetation canopies and map underlying structures. Since then, significant improvements in technology have resulted in lidar becoming an essential exploratory tool for archaeologists worldwide.

Great book on LIDAR & South America:

The Lost City of the Monkey God: A True Story" by by Douglas Preston

Some additional reading on new discoveries over the past few years:

Lidar Shows Huge Network of Ancient Cities in the Amazon | Moss and Fog

Recent lidar mapping of deep jungle show vast areas of ancient human civilization in the Amazon.

mossandfog.com

Ancient cities discovered in the Amazon are the largest yet found

A mysterious civilisation built a network of cities and roads in the Amazon between 3000 and 1500 years ago, and then disappeared

www.newscientist.com

More than 100 archaeological structures discovered in the Peruvian Andes

Built by the Chachapoya civilisation, known as the “people of the cloud forest”, the structures are located within a Unesco Mixed World Heritage site 500km north of Lima

www.theartnewspaper.com

The Lost City That’s Not Lost, Not a City, and Doesn’t Need to Be Discovered

Modern explorers can “discover” an ancient site, but the people living in the area already have deep knowledge about their region’s history.

www.sapiens.org

LiDAR technology shines new light on Guatemala's ancient Maya ruins | Washington Diplomat

Guatemala, Central America’s largest nation, often makes headlines, though nearly always for the wrong reasons—violent crime, drug trafficking, natural disasters and illegal immigration. When it comes to uncovering ancient civilizations, however, Guatemala is a world leader.

washdiplomat.com

Data from various sites throughout South America:

Research Just Showed That The Maya Population Was Much Larger Than Experts Thought And May Have Included 16 Million People

Researchers used LiDAR scans of Maya urban centers to conclude that, when the population was at its peak circa 600-900 C.E., this civilization was much larger than experts once thought.

“This discovery has proven there was an equivalent of Rome in Amazonia,” Rostain said. “The people living in these societies weren’t semi-nomadic people lost in the rainforest looking for food. They weren’t the small tribes of the Amazon we know today. They were highly specialised people: earthmovers, engineers, farmers, fishermen, priests, chiefs or kings. It was a stratified society, a specialised society, so there is certainly something of Rome."

...

"“Using airborne laser-scanning technology (Lidar), Rostain and his colleagues discovered a long-lost network of cities extending across 300sq km in the Ecuadorean Amazon, complete with plazas, ceremonial sites, drainage canals and roads that were built 2,500 years ago and had remained hidden for thousands of years."

But LiDAR, said Estrada-Belli, “has revolutionized our ability to map.” The technology has enabled archaeologists to cover around 7,000 square kilometers as of 2019, and to recognize virtually every structure, “even small things you couldn’t see even if you were standing right in front of it—but also very large things because their size is obscured by the jungle itself.”

In just six months, archaeologists have managed to scan an area ten times larger than what five years’ worth of standard pedestrian surveys had covered. Technology can help collect data and obtain the variations, elevations, and mysteries behind the Mayan landscape, such as information about population density and cultivation practices.

In addition, the data can be presented to appeal to a wide variety of audiences; day and night views, 3-D enhancement, and thermal panoramas are just a few of the filters users can choose to help make the data come to life.

 

[Kaido]

3D retexturing!

https://twitter.com/x/status/1977457657876226446