News for hackers. Stuff that matters.

Show HN: A decent text-to-image model trained for ~$20K (Apache 2.0 Weights)

https://www.linum.ai/field-notes/pjit-21k

My brother and I run a 2-person lab that trains our own generative video models from scratch. The ultimate goal is a set of creative tools that make it easier for anyone to make animation.

Video models today are too slow/expensive and lack sufficient controls. Creativity is an iterative process. You make, you tweak, you get better. In our opinion, they're not usable yet for storytelling.

When we set out to build Linum v3 (our next video model) earlier this year, the primary goal was to bring the cost envelope of inference down significantly. If our model is fast and cheap, it's the first step (of many) towards useful creative tools.

Typically, you see "latent diffusion models". A VAE (Variational Autoencoder) compresses the input data into a smaller "latent space" and the diffusion model learns how to generate within this subspace.

Over the past few months, we've been publishing our work on efficient pixel-space diffusion models. These throw away the VAE, having the diffusion model learn a latent-space optimized for its exact task.

This serves 3 purposes: 1) Simplify Architecture (no 2-stage model training) 2) Faster Convergence (11x faster convergence vs. our latent diffusion counterpart) 3) Cheaper Inference (4x more aggressive token reduction in attention context-windows, without significant loss in quality)

This past month we were pressure testing our latest architecture (Pyramid-JiT) by training a text-to-image model longer than the typical one-off experiment. Specifically, we trained it for ~$20K on 660M samples.

We were surprised by the quality of results, so we thought we'd release it as open weights for the community (Apache 2.0). The research itself is already public, on our website.

For us, it's a big stepping stone towards cheaper video models, since pretty much all video models start as image models. We like to say these models have to learn nouns before they learn verbs.

( 29 comments | 2 points | on HN )
Re:Show HN: A decent text-to-image model trained for ~$20K (... by gigel82 on Thursday November 05, 2026 @12:15PM (#50381665)
I wonder how this behaves when quantized, and how far are we from being able to run something like this purely in the browser with WebGPU.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by Valk3_ on Thursday November 05, 2026 @12:29PM (#50381842)
What was the most difficult challenge for you guys with this project?

Also I'm a little intrigued on looking into efficiency techniques for video gen. What are current research directions in that field and promising directions you are particularly interested in? Any pointers for beginners?

Re:Show HN: A decent text-to-image model trained for ~$20K (... by yhjc2692 on Thursday November 05, 2026 @04:17PM (#50384571)
I think there's a whole plethora of research on efficient attention implementations (similar to how DeepSeek & Kimi has made strides with theirs in LLMs) [1], [2]. And various techniques to reduce the flow-matching / diffusion iterations ("reflow") [3].

CVPR 2026 paper list (after filtering on video generation) might be a good start - asking LLMs to take a look and filter relevant papers sound even better.

[1] https://arxiv.org/abs/2609.15810 [2] https://arxiv.org/abs/2602.13515 [3] https://arxiv.org/abs/2605.23381

Re:Show HN: A decent text-to-image model trained for ~$20K (... by Valk3_ on Friday November 06, 2026 @05:10AM (#50390447)
Thanks! And sorry, I was just curious to get some opinions on it from people who have worked with that kind of stuff. I feel like someone who has worked on it in a serious capacity will be a better filter than LLM (I don't know enough to provide the right keywords that would give me high signal papers). However, I realized I should probably just read a survey to get some overview on the state if I can't get any answers.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by gunalx on Thursday November 05, 2026 @01:03PM (#50382209)
Interesting. It does however seem to struggle with a lot of the typical t2i models struggle with. Over smoothing, lots of uncanney shine, mangled limbs and such.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by dvh on Thursday November 05, 2026 @01:22PM (#50382483)
How many ladybugs are in training dataset?
Re:Show HN: A decent text-to-image model trained for ~$20K (... by whythismatters on Thursday November 05, 2026 @01:42PM (#50382762)
Re:Show HN: A decent text-to-image model trained for ~$20K (... by tensor on Thursday November 05, 2026 @02:10PM (#50383078)
Amazing. I think this is an understudied areas. The big labs if anything are disincentivized to make efficient models as currently access to compute is one of their moats. That's not the same for Chinese labs, which is why you see groundbreaking efficiency work mostly coming from China.

Really happy to see a small shop like yours doing this sort of work!

Re:Show HN: A decent text-to-image model trained for ~$20K (... by blueblazin on Thursday November 05, 2026 @10:13PM (#50387715)
That makes no sense. The big labs don’t reveal their efficiency research but that doesn’t mean they aren’t doing it. They can charge the same prices while making models more efficient internally and increase margins…
Re:Show HN: A decent text-to-image model trained for ~$20K (... by darkwater on Friday November 06, 2026 @02:19AM (#50389203)
I bet they do that as well but I also think that if they have to choose between inference speed or "intelligence" they would choose the latter.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by yorwba on Friday November 06, 2026 @05:41AM (#50390689)
They don't have to choose, though, since they can release multiple models with different tradeoffs.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by darkwater on Friday November 06, 2026 @10:52AM (#50393898)
This would imply they have infinite development capacity
Re:Show HN: A decent text-to-image model trained for ~$20K (... by yorwba on Friday November 06, 2026 @01:23PM (#50396204)
Not infinite, just enough to make more than one model. OpenAI and Anthropic both have four different size categories...
Re:Show HN: A decent text-to-image model trained for ~$20K (... by Robin_Message on Thursday November 05, 2026 @03:02PM (#50383724)
What are you using as your training data out of interest? Very impressive generation.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by jcattle on Friday November 06, 2026 @09:03AM (#50392411)
I'm always wondering that. Where do people get 660 million labelled images from?
Re:Show HN: A decent text-to-image model trained for ~$20K (... by robotswantdata on Thursday November 05, 2026 @03:13PM (#50383865)
What open source datasets did you use? Or did you pirate like the big labs ?
Re:Show HN: A decent text-to-image model trained for ~$20K (... by micw on Thursday November 05, 2026 @03:18PM (#50383926)
Impressive. Even the average number of fingers seems to be correct ;-) Looking at the failures, I wonder how much cherry picking went into the showcase.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by yhjc2692 on Thursday November 05, 2026 @04:10PM (#50384496)
Wow this is really impressive. Throwing away the VAE feels counterintuitive - I'm shocked it works so well! Appreciate that you shared your training insights on the blog post!
Re:Show HN: A decent text-to-image model trained for ~$20K (... by jcattle on Friday November 06, 2026 @09:01AM (#50392394)
$100k seems like a lot. I guess it isn't in the grand scheme of things but still
Re:Show HN: A decent text-to-image model trained for ~$20K (... by Tarq0n on Friday November 06, 2026 @10:38AM (#50393665)
Does it? Text generation models have been decoder-only for a while, with great benefits to their training. It's surprising to me that more attempts weren't made for other modalities.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by fzysingularity on Thursday November 05, 2026 @05:19PM (#50385303)
Congrats guys, this is super cool. I think if you built an arena where you can drive the cost even lower ($1K), it'd really get developers stoked.

Have you thought about setting up auto-research on this for smaller experiments? What's the minimum-viable budget for these types of workloads?

Re:Show HN: A decent text-to-image model trained for ~$20K (... by mkl on Thursday November 05, 2026 @05:32PM (#50385454)
It looks really good! Except almost every hand is wrong. I'm guessing that's a training data thing? Can you tell us more about the 660M training images and prompts? Where did they come from and what do they cover? (I've seen the README say text was excluded.)
Re:Show HN: A decent text-to-image model trained for ~$20K (... by sebion7125 on Thursday November 05, 2026 @06:11PM (#50385887)
This is awesome. Thanks for sharing weights on HF :-)
Re:Show HN: A decent text-to-image model trained for ~$20K (... by jjcm on Thursday November 05, 2026 @11:07PM (#50388054)
Very nice all. I've been playing around having poor results of training a pixel space image model of my own, though in a space very opposite from what you're doing (UI design, very heavy on text).

Any chance you'll release your training recipe as well? I know the model code is public, but I'd love to dive into the full process.

Re:Show HN: A decent text-to-image model trained for ~$20K (... by andai on Friday November 06, 2026 @01:49AM (#50389020)
Well done. $20k? That's insane. I remember tinyllama back in the day (2023) was trained on $40k and could barely form coherent sentences.

Are pictures easier than English, or did the methods advance so much?

Re:Show HN: A decent text-to-image model trained for ~$20K (... by wongarsu on Friday November 06, 2026 @07:27AM (#50391558)
Over the last 3 years we made a lot of progress in efficient training methods. On text models, you can now train GPT-2 levels of coherent English from scratch for about $50 (e.g. using Karpathy's nanochat [1])

1: https://github.com/karpathy/nanochat

Re:Show HN: A decent text-to-image model trained for ~$20K (... by andai on Friday November 06, 2026 @10:19AM (#50393378)
That's insane. How does that work? Does it take so much less data now, or smaller models, or what? Apparently GPT-2 cost ~$40,000 to train when it came out.
Re:Show HN: A decent text-to-image model trained for ~$20K (... by pruthvishetty on Friday November 06, 2026 @04:16AM (#50390030)
Bravo!
Re:Show HN: A decent text-to-image model trained for ~$20K (... by axiom_01 on Monday November 09, 2026 @12:36PM (#50428733)
Really impressive work, especially with such a small team and budget. Removing the VAE while improving efficiency is an interesting approach. Curious to see how this architecture scales to video generation!