Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge

https://qwen.ai/blog?id=qwen-image-3.0

by ilreb • 2 days ago

564 213 2 days ago

qwen.ai

xiaoyu2006• 2 days ago

The blog write-up style is so casual haha.

embedding-shape• 2 days ago

Not a single word about when/if they'll actually release the weights for this, or am I missing it somewhere?

user439282• 2 days ago

Considering Qwen-Image-2.0 weights have not been released either, it unfortunately looks unlikely.

vachina• 1 day ago

Why release it just so Cursor/Azure/Amazon can profit off of it? Unless OpenAI actually opens it up fat chance.

vunderba3• 1 day ago

After Z-Image and the original Qwen-Image - I think they've pivoted completely to closed source. From the images I've seen, Qwen-Image 3.0 is just a subpar equivalent to other proprietary models like gpt-image-2 and nb-pro.

Mashimo3• 2 days ago

Supports up to 4.5k token input, effortlessly generating complex layouts such as newspapers, storyboards, and exam papers.

Impressive.

Btw, what is currently the best model to run locally on a 16GB Vram? Is it Z-Image Turbo?

woadwarrior013• 2 days ago

Krea-2-Turbo. I've even got it working locally on my M5 iPad Pro.

Mashimo3• 2 days ago

thanks mate. Sadly not supported yet by Invoke, but I will take a look.

Izmaki2• 1 day ago

We implemented safety measures across the full model development lifecycle.

Any suggestions for the best open, non-opinionated model?

kroaton• 24 hours ago

It depends on what you need, but Krea/Klein9b/Ideogram4/Z-Image are among the best right now for text2image and Qwen Edit and Klein are probably still the best at editing.

spwa42• 2 days ago

Appears to be closed-weights entirely. No word at all on any weights release.

weird-eye-issue5• 2 days ago

The meta keywords in the HTML is very interesting. 100+ references to NSFW topics such as hentai, nudes, etc.

walrus016• 2 days ago

Wow you really weren't kidding. Fully automated AI slop tentacles for everyone!

https://pastes.io/uenL6X9K

It also seems to have an obsession with this celebrity, based on how many times ctrl-f for "stefani" turns up a result.

https://en.wikipedia.org/wiki/Gwen_Stefani

Mashimo3• 2 days ago

total drama qwen hentai

Do you want to know more?

[ ] Yes [x] no

weird-eye-issue5• 2 days ago

spider qwen thicc, spider qwen tied up, spider qwen toddler costume, spider qwen trans, spider qwen wallpaper, spider qwen xxx hentai

WithinReason• 2 days ago

probably due to misspellings as qwen stefani

alex_suzuki• 2 days ago

ben 10 hentai game where qwen gets drunk

erm... what?

pdpi• 1 day ago

Keyword typosquatting I guess?

goobatrooba• 1 day ago

Wow, crazy things - "shemale porn", "spider fucks venom", and various porn URLs like tushy.com. What the hell did they do to their AI? Is that the training corpus or intended use case?

chmod775• 2 days ago

Bonus points for also covering misspellings like "pregbant".

QuantumNomad_• 2 days ago

how is prangent formed

Am i gregnant?

https://youtu.be/EShUeudtaFg

isoprophlex• 2 days ago

show bobs and vegana!

lukan3• 2 days ago

That is hilarious as porn is illegal in China. But I guess pron SEO is allright, if it is against the west.

weird-eye-issue5• 2 days ago

Western search engines simply ignore the meta keywords tag for over a decade now

vitalyan8184• 1 day ago

uh, are you implying that porn is bad somehow? I can show you a hundred articles that say only nazi incel chuds believe so.

https://www.nytimes.com/2019/06/07/us/hate-groups-porn-consp...

porn is as American as apple pie.

postalcoder4• 2 days ago

That's hilarious. These keywords are applied globally, even on pages like https://qwen.ai/usagepolicy, where they very kindly ask you not to use their products for sexual content.

Run this in console to see all the tags:

  document.querySelector('meta[name="keywords"]').content

khurs2• 1 day ago

WTF lol

    who made qwen stefani's dress

    spiderman into the spider verse did qwen meet Peter

    qwen stacey porn

    do blake shelton and qwen stafani have children together

LoveMortuus• 1 day ago

Why would they added URLs for porn websites to their keywords? Like literally the names of websites with ".com" and such at the end.

WhereIsTheTruth• 1 day ago

Brainrot is the new opium war

cyanydeez• 1 day ago

how the turn tables

user439282• 1 day ago

I assume they use a SEO tool that automatically adds these meta keywords to optimize for some search engines.

It apparently adds common search terms that contain words like "qwen". This evidently includes possibly mistyped searches for "gwen" or "ben" in a NSFW context.

Maybe someone knows more about how such SEO tools work, and where they pull the data from.

spiderfarmer• 1 day ago

Sounds like “keyword shitter” (genuine tool)

01284a7e• 1 day ago

What value does the 'keywords' meta tag even have these days? 77+ KB of crap added to the page weight. Web development is full of idiots.

j0ej0ej0e• 1 day ago

Maybe it's completely intentional and it's making a comeback at least with LLMs? Google say it has no value, but LLMs probably genuinely use it.

drakythe• 1 day ago

Interesting too that they would include "ai friend" when China just added restrictions on AI "partners" such that many services stopped offering them rather than try to adhere to the restrictions.

ssalka• 1 day ago

I think the Qwen team is probably aware that the NSFW community is very quick to adopt any new image gen model (see: Civitai). So, it seems like a good SEO approach to try and surface their model in search results that said community is likely already checking.

rvz• 2 days ago

Midjourney already knew that image generation was going to zero. Again yet another reason why the model was never a moat in the first place.

amelius• 1 day ago

The moat is the training data. But somehow we've collectively decided that it's not.

treetalker• 2 days ago

The red-dress woman's vestigial pinkie toes …

postalcoder4• 2 days ago

They must have trained on GPT Image 1 outputs. The yellow tint is unmistakable.

https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen-Image/i...

yorwba3• 2 days ago

GPT Image 1 ended up with a yellow tint without training on another image-generation model's output. It's just that humans like pictures with a soft sunset glow, and this is a very easy global signal for a preference model to pick up on, and for a image-generation model to imitate. So optimizing for aesthetic appeal makes everything slightly tinted by default, unless you make sure to countersteer.

postalcoder4• 2 days ago

Fair point!

Grimblewald• 2 days ago

Didnt the piss tint come after their big studio ghibli heist?

dannyw• 1 day ago

AI-generated images are part of the web now, if you're doing ordinary web scraping, you can't avoid training on generated images.

hugmynutus• 1 day ago

yellow/red tint is an extremely common problem not matter the photograph source you train on

source: work at a photograph start, even training on raw images things get tinted, it is an uphill battle

greyb• 1 day ago

You clearly haven't met a Chinese RedNote user.

cousinbryce• 1 day ago

Maybe it was trained in a Mexican data center https://en.wikipedia.org/wiki/Mexican_filter

porphyra• 23 hours ago

Foreign media like the BBC were actually caught applying a yellow filter to footage of China [1] [2] so maybe they trained on that lol

[1] https://www.globaltimes.cn/page/202102/1215710.shtml

[2] https://xcancel.com/thouse_opinions/status/13616465112479539...

jdw642• 2 days ago

Wow, it displays Korean properly without breaking. But there are still a lot of typos. Haha, it's good that Korean displays properly, but there are a lot of incorrect sentences

sheept• 2 days ago

I find it mildly interesting how your comment repeats itself but with different phrasing

jdw642• 1 day ago

It's because of the structure of Korean. I think in Korean first, so I end up translating it directly.

When I want to emphasize something, I tend to repeat it

mahimai• 2 days ago

interesting

mynti• 2 days ago

To me it feels so weird that people are trying to push these model for online shopping like "here is how this dress/shirt/pants would look on you". But these models will always make the clothes fit your body and show you in flattering light and so on. How the actual garment fits is still as elusive as before these tools

walrus016• 2 days ago

The cynic in me says this is just an advancement and logical continuation from the known problem of sycophantic behavior in text-to-text chatbot format LLM, to image generation models.

inerte• 1 day ago

To me the tools are newer, but fit people in a controlled context to sell clothing is as old as photography itself.

number6• 1 day ago

Sounds exactly like something the marketing department would use to up the convertion rate

viraptor2• 1 day ago

There's definitely a predatory "everything will look good", but you can also leverage exactly that part for your own purposes - I've done pre-shopping a couple of times by asking for something like "a grid of 9 versions of my photo, wearing different types of X that look good". Definitely helped with choosing a good style.

khurs2• 1 day ago

As long as the Product Manger gets a bonus/promotion then it's a success.

TylerE• 1 day ago

To be fair in many cases the actual clothes aren't much better. I've had two pair of the same pants, same brand, same size fit noticeably differently.

morningsam• 1 day ago

Maybe this will lead to more business for tailors doing alterations, assuming the clothes people end up buying are expensive enough to justify it.

jimnotgym• 1 day ago

I understand they are already doing decent business since the advent of the magic weight loss pen

epolanski• 1 day ago

I have a similar use case at work for previewing construction material and such in our catalogue applied to user uploaded images.

Results are mixed, expensive, but it really feels you're few months off the next improvement to really nail it. It's already good enough.

Wonder what Qwen image will provide over nano banana.

vunderba3• 1 day ago

Well from the images I've seen they seem to be pushing text-heavy "infograph" capabilities pretty hard. Outside of that, I have serious doubts that it'll be better than other proprietary models like NB Pro and gpt-image-2.

k2enemy• 1 day ago

I've noticed something similar in Facebook marketplace ads for used furniture. Most of the images are AI generated to look like a Pottery Barn catalog, then the last image will be the actual item, full of scratches and other damage, sitting in a messy garage.

aembleton• 1 day ago

Estate agents are also doing it; making interiors of houses look very different to how they really are.

QuantumGood• 1 day ago

And knocking out the messy garage background to replace it with a showroom is the easiest prompt.

mikepurvis2• 1 day ago

I've started to see this on Etsy and Wayfair too, where there will be a listing that is clearly just MDF flatpack being resold from China, but the AI-generated images wildly exaggerate the proportions of it.

Here's a recent example: https://www.etsy.com/ca/listing/4509158065/corner-wall-shelf...

Ironically, ChatGPT is decently good at ferreting these out. Like I sent it a screenshot of that listing and it not only helped me find where the original item was for sale, but also pointed out how the dimensioned diagram shows it as being just 49" tall, whereas the "in real life" image looks like it's at least six feet, based on it coming up over the top of the picture frame.

I ended up engaging a local woodworker to make me a piece like it instead. Obviously an order of magnitude difference in price, but it will actually be real solid walnut and finished to match my dining table.

jliptzin• 1 day ago

If it doesn’t fit, then you must have gained weight while the item was in transit

agile-gift02622• 1 day ago

You can fix it through our partner, Ozempic

jareklupinski• 1 day ago

i'd like it to notice things i would miss, like "this is ring-spun shirt, so it will sit like this on your torso" or "these pleats will require you to iron them" etc

pwillia7• 1 day ago

I bet if you had a way to easily train a LORA on you trying on various clothing types the models would do pretty well but I agree without that I can't think of a way to get it to work. Reminds me of the flattering mirrors scam https://finance.yahoo.com/news/company-swears-controversial-...

brookst• 1 day ago

The more honest ones are “how it looks on you” and don’t promise fit that depends on so many measurements that aren’t even visible in pics.

Sure, it’s idealized, but some people benefit from seeing color / neckline / etc on themselves as a visual reference.

Me, I’m a text-learner so I don’t get it at all. But I know people who get value.

therealpygon2• 1 day ago

Good or bad, I’m pretty sure showing things in the best light has always been the point of “marketing”. There is a reason ads aren’t filled with ugly people with misshaped bodies, and it isn’t because the intention is to reflect reality, so what you allude to as problematic is what a lot of businesses call a feature.

sixothree2• 1 day ago

I'm pretty sure I saw somewhere people can sell re-sell clothes. They can include a picture of themselves wearing the item and it replaces literally everything but the clothing with someone/someplace "prettier".

In the end the one thing that is completely honest is the portion of the picture that is the item you are re-selling. But somehow to me the entire thing feels disingenuous.

LogicFailsMe• 1 day ago

I use nanobanana 2 to test changes in paint and flooring/tiling to great success. And the real pro move is taking those images to a designer to tweak the remaining 20% or so.

sixothree2• 1 day ago

You just described how people are using AI in general. And I think that last step is actually a huge gap. People will brainstorm or improve their understanding of the task with AI and create something. Ideally they would hand it off to a professional for finalization like you did.

I wish a service existed where I could make something in AI (text, images, whatever), then pass that AI output to an actual human who would use it as a guide to produce an actual product.

paradox460• 1 day ago

I used it recently to explore a few placements of a pool and glass building on my property. It's not authoritative, by any means, but it did answer some questions

And at this level it's barely any different than an architectural render, more for "how could this look" rather than "how will this look"

qwertox• 1 day ago

But these models will always make the clothes fit your body and show you in flattering light and so on

This is something that can be fixed over time. And if this forces clothing manufacturers to stick more to their advertised "specs" (width/length), then it's a win for us.

coldtea2• 1 day ago

Nobody that's pushing it has any incentive for fixing it - and they have all the opposite incentives.

victorbjorklund• 1 day ago

Is it gonna be less distorted than just seeing the shirt on a model photographed by a professional in the perfect light?

collinmcnulty• 1 day ago

Yes, because it is the actual dimensions of the real shirt.

smith7018• 1 day ago

Many product/model shots actually use clips to make the clothing look like it's perfectly fitted to the body. [1] This image is actually over 15 years old at this point. I think there should be laws that prevent this because it veers into false advertising though others believe it's alright because you can theoretically tailor the clothes to fit like this.

Regardless, I'd rather see real clothing on a real person when it comes to my purchasing decisions. I buy a lot of vintage clothes online and I've noticed a dramatic uptick in AI images of models wearing the clothes. I've never once bought from those sellers because it feels disingenuous. Sometimes they have fake runways which is actual false advertising because it makes the item appear more expensive than it really is. I've also noticed that the AI models' body types are always thin even if the item is a L or XL. Needless to say, the AI isn't showing me what an XL looks like on a small model; it's showing what a small model would look like if the item fit perfectly.

[1] https://www.primermagazine.com/wp-content/uploads/2011/02/St...

coldtea2• 1 day ago

Yes, because at least that won't be flattering you (not to mention it would be showing the actual garment).

spaceman_20202• 1 day ago

A prompt went viral recently where people were sharing their pictures and asking chatgpt to visualise what their looksmatched partner would look like

The result was always someone extremely good looking

There’s going to be an entirely new class of mental disorders that will emerge from people being deluded by AI

zkmon• 1 day ago

How does this fry pan look with a fish in it? Ask Mr Bean.

teraflop• 1 day ago

The short-term goal of a tool like this is to sell products. The more ambitious long-term goal is to shift cultural norms, blurring the lines between advertising and reality until the question you're asking is no longer consciously asked. At least, not by average people, and not at the point of purchase.

I find it easy to envision a world, maybe 50 years from now, in which the very concept of "truth in advertising" is viewed as a lost, idyllic fantasy. Something people are nostalgic for, but feel powerless to regain.

dataflow• 1 day ago

I find it easy to envision a world, maybe 50 years from now, in which the very concept of "truth in advertising" is viewed as a lost, idyllic fantasy.

From what I hear (fact-check, etc. needed) that's already the case right now with much more consequential transactions, like renting real estate in NYC. Square footages that are blatant lies, etc.

verisimi• 1 day ago

I find it easy to envision a world, maybe 50 years from now, in which the very concept of "truth in advertising" is viewed as a lost, idyllic fantasy.

I find it easy now!

palmotea• 1 day ago

I find it easy to envision a world, maybe 50 years from now, in which the very concept of "truth in advertising" is viewed as a lost, idyllic fantasy. Something people are nostalgic for, but feel powerless to regain.

It's infuriating the amount of effort people will expend to claim that what was achieved in the past is literally impossible to do now. It's pervasive, especially from allegedly-smart people like software engineers.

ElProlactin• 1 day ago

I find it easy to envision a world, maybe 50 years from now, in which the very concept of "truth in advertising" is viewed as a lost, idyllic fantasy.

There hasn't been "truth in advertising" for many, many years. The only thing that has changed recently is that you don't even have to hide most of the lies.

supern0va2• 1 day ago

I find it easy to envision a world, maybe 50 years from now, in which the very concept of "truth in advertising" is viewed as a lost, idyllic fantasy. Something people are nostalgic for, but feel powerless to regain.

Alternatively, perhaps we'll see models fine-tuned or steer-able towards accuracy that customers can themselves use to get a more honest view on what the product would look like in person.

The funny thing about these tools is that they can go either direction, but it sure seems like there's the potential for it empower individuals and shift the balance. Clothing sales shifting online has given sellers the advantage/ease to deceive without much customers can do other than hope the reviews aren't manipulated (they are) even before AI. Maybe this can turn things around as people start to shop with personal agents.

newswasboring• 1 day ago

I've been struggling with this question myself. But isn't this a model training/use problem (a.k.a skill issue )? Isn't there a way to make these models be faithful to how people will actually look?

daedrdev• 1 day ago

But they dont have to do either of these things. Thats just what people prompt

aljgz• 1 day ago

One joy of online shopping, especially for people doing it in an impulsive and/or addictive way (it's not really rare) is the satisfaction they get from the imagination of having it.

An acquaintance of mine was buying many and not wearing most, as she did not attend that many social occasions. Still, she kept buying.

Eventually, she had to face the actual problem in her life that bothered her. She ended up dealing with it, terribly.

geniium• 1 day ago

Yet it pushes the goal of the brand/shop forward: make the product more appealing and sell

sandworm101• 1 day ago

User-targeted fashion advice is the polite goal. AI designed to generate images of people is actually racing to capture the entertainment markets. They want to be ready to replace models/actors in everything from fashion mags to porn studios. That is where the money is.

yorwba3• 1 day ago

For Alibaba, it's definitely all about shopping. That's where their money is. They haven't had much luck investing in entertainment so far.

jeromechoo• 1 day ago

It’s clear to us here on HN but to the average person today the way these models actually work is beyond the realm of constraints and reasoning.

imhoguy• 1 day ago

This will push us even more to go outside and visit actual shops. Becouse thanks to AI we will have more time? Will we?

kmfrk• 1 day ago

Definitely. I think the appeal to some isn't GenAI per se but image-to-image generation like "make this item look like a talented photographer shot it".

But I can't imagine that your conversion wouldn't benefit by showing actual imperfect photos of your products, especially when you run a platform with 1k new listings per month. I'm personally turned off by something that looks like 3D, especially when all the different colour variations look the exact same.

m463• 20 hours ago

I see the ai product ads and it's pretty funny. The product is shown but is "broken" as far as scale is concerned.

Sort of like how they tried to make gandalf big and the hobbits small in lord of the rings (which wasn't very convincing to me)

DrScientist• 10 hours ago

If the online retailer has a no quibble returns policy - then at there is at least a strong incentive to minimise returns, rather than oversell.

viridir• 2 days ago

This looks impressive！

pal9000i• 2 days ago

How long until we get rid of the AI "plasticness" in portrait kind of generated images?

aitchnyu• 1 day ago

There is an IRL phenomena, glass skin skincare and makeup, where the person's skin is evenly flat and toned and glossy. Did you think they are more plasticky than IRL models?

numpad0• 1 day ago

That's supposed to make skin appear lively. Human skins in AI images tend to look clouded, opaque, and overall un-alive, so to speak.

jrs100000• 1 day ago

The right models and LORAs can get rid of it right now. People apparently really like everyone to look like over exposed over filtered mannequins, so the big companies target that look.

cubefox2• 1 day ago

I think it's unintentional. People usually dislike any recognizable "AI look", but they do like other aspects which might have an unrealistic AI look as a side effect.

For example, Google's Imagen 3 usually looked a lot less fake than the newer Imagen 4, but the latter still scored higher on most benchmarks because it made fewer mistakes and had better prompt following capabilities.

A similar thing happened with Dalle-E 2 and Dell-E 3: The new model was better but also more fake looking.

lifeofpi331144• 1 day ago

what are the right models and loras?

saltysalt• 2 days ago

It will be interesting to compare this to Flux 2.

Oarch• 2 days ago

I assume Van Gogh didn't paint enough hands to train from!

gpjanik2• 1 day ago

The real performance is nowhere close to what is presented in the marketing materials, which is pretty annoying. Especially text rendering and accuracy.

Try asking it for a plot of Polish GDP growth over the past 20 years. It's slop.

viraptor2• 1 day ago

This model isn't supposed to contain all the numerical data. It will give you a (usually) matching graph transformed from one you provide or from a table of information you provide. Or you can pipeline from an LLM doing research on that data first. But expecting an image gen model to get you GDP info has got to be one of the worst possible approaches.

Especially text rendering

That's true though. I still got some completely fried letters in headings.

yorwba3• 1 day ago

Giving it numerical data in an LLM-generated prompt doesn't seem to help much: https://imgur.com/a/KFhczOd

It included the table verbatim and even managed to hallucinate a reasonable heading for it, but then the graph doesn't even manage to align the data points with the time axis, leading to an unfortunate collision in the middle.

I guess you should use a traditional graphing library for your presentation slides for now.

gpjanik2• 1 day ago

I am not expecting it, it's the Qwen team is claimingthey can do much harder tasks than this, like rendering a consistent page of a maths paper, or creating true to fact explainers.

They can't.

dsrtslnd23• 1 day ago

Seems that this will not be open weights?

gchokov• 1 day ago

It failed to create a simple overlay on a map - something ChatGPT had no issues with.

maxloh• 1 day ago

I am curious whether the model requires a font to be installed. Does it also generate the glyphs for the text?

simonw2• 1 day ago

Yes, it generates the text without using a font. Same is true of other image models like ChatGPT Images and Gemini Nano Banana and Midjourney.

vunderba3• 1 day ago

It's a proprietary model as a service - it doesn't require anything outside of a browser. But even when they released open-weight (the original Qwen-Image) it's a diffusion model and can't use custom font files.

feverzsj• 1 day ago

The "piss filter" is still everywhere.

dhbradshaw• 1 day ago

The generated latex pdf!

hessammehr• 1 day ago

Very cool but the Arabic text in the title image is obviously and hopelessly broken, which is oddly not the case when actually using the model. Could it be that the hero image was not generated by Qwen Image 3.0?

amrrs• 1 day ago

that's a very interesting point to test new models. I speak an Indian language called "Tamil" and have always tested new models with Tamil but also have tried little bit of Arabic (quranic verses) with previous GPT image 2 and Nanobanana pro and they have nailed it. Don't know if it was because of extensive training data.

bejd• 1 day ago

I wonder if they got permission to generate that (admittedly impressive) Berserk image.

arslan9063• 1 day ago

THIS IS EXACTLY WHAT I WAS LOOKING FOR

ninjagoo• 1 day ago

The examples posted on their launch blog page are quite impressive, especially for fine details, multi-panel/multi-page and text rendering.

But: not open-source/open-weights, and no indication that weights/source will be released either.

simonw2• 1 day ago

to precisely describe the full 3×3 grid takes a full 3.7k tokens

It's a shame they didn't share that prompt - it would make that demo more convincing.

tarcon• 1 day ago

I am surprised by the rather bad output. It doesn't achieve qwen image 1 quality in composition or anatomical correctness. Tested on chat.qwen.ai

I am seeing third legs and glowing eyes. It's a Microsoft Lens level of quality and that one was pulled.

sajithdilshan2• 1 day ago

I truly wish these models were available when I was in University. As a visual learner it would have been much easier for me to understand certain topics with illustrative diagrams rather than reading a wall of text.

DaiPlusPlus2• 1 day ago

…but the illustrative diagrams are a simulacrum; if you ask Qwen, or any image-generator, for an “accurate” poster-design featuring a representation of a model of an atom and explaining its constituent parts I expect you’ll get an imitation-airbrush rendering of red, blue, and grey table-tennis balls orbiting in perfect circles; you might get an electron-shell diagram if you’re lucky. What you won’t get is anything remotely related to probability-clouds.

Edit: just to test myself I asked Nano Banana 2 to generate “an undergraduate infographic poster about how atoms work” - and the result was something right out of a middle-school science textbook and very Bohr…

sajithdilshan2• 1 day ago

This is the image I got for the same prompt: https://jumpshare.com/s/mqqBdl7U59FWPiEXjwoM. It's more like high-school level, but not bad. I can imagine a collage professor can improve the prompt the create a more accurate and detailed diagram

zarzavat• 1 day ago

This seems like something that could be solved by asking an LLM to write the prompt for the image model. You can also feed in the output of an image model into an LLM and ask it to check it/make improvements.

wincy• 1 day ago

My wife has taken all her recipes and fed them through ChatGPT image gen to make zine pages and they’re really cool! She’s building a recipe book for the kids so they’ll know all the recipes from their childhood.

lifthrasiir• 1 day ago

In the three examples below, the model accurately renders Japanese, Korean, and Spanish respectively.

And yet the Korean text is not accurate... [1]

[1] E.g. "드레스 컬렉션 dress collection" has vowels ㅔ mixed with ㅐ, "초웜한" should be "초월한 exceeding", "신키한" should be "실키한 silky", "디자언되다" should be "디자인되다 have been designed", "로얼" should be "로열 royal", and so on.

ndom91• 1 day ago

Again not released on huggingface immediately?

timedude2• 1 day ago

Zooming in on mobile on that website causes a large white area to obstruct the page. Might wanna look into that.

As for the image model, wow...

aand16• 7 hours ago

Their site is still broken, months later. Pinching to zoom in progressively covers the content on the page.

Kinda wild to think they haven't fixed it yet. On mobile you cannot zoom in to view their images.

Maybe the site is vibe coded and nobody at Qwen actually knows how to fix it..

Another astonishing UX fail was (still is) on Xiaomi's page[1] - where they try, and successfully fail, to showcase a video editor implemented by their model.

They embed a screen capture hosted on a Chinese hosting site BiliBili, too bad the video is 360p pixel soup.

To be able to change the resolution you are required to:

- know to use desktop view in the browser to make the resolution button show

- know Chinese to understand which button is "resolution" and that you need to login to switch resolution

- have and account with the video hoster or be willing to register

I was curious, but not that curious...

[1] https://mimo.xiaomi.com/mimo-v2-5-pro#AFull-FeaturedVideoEdi...

Smaily• 1 day ago

why I cant submitte new posts here ? My account since 2016

cubefox2• 1 day ago

Probably because you have too few upvoted comments. The rules likely got stricter when LLMs started to infiltrate Hacker News.

luciana1u• 1 day ago

the natural endpoint of this technology is product photos that look better than the actual product, which is going to make unboxing videos the last remaining source of truth on the internet

jcattle• 1 day ago

What I can not wrap my head around: How are these models trained?

What training mechanism or model architecture provides the glue to go from human text to images?

Don't you need to have millions of really descriptively labelled images?

mistercheph• 1 day ago

I found this 3b1b guest video on diffusion helpful: https://www.youtube.com/watch?v=iv-5mZ_9CPY

nucleative• 1 day ago

That's exactly how they do it.

There are ML models that do the reverse and output image to text, which assist quite a lot.

The better the text represents the unique thing in the photo, the better the model understands what that text means.

vonneumannstan• 1 day ago

Short answer yes.

Slightly longer answer for older text to image models you teach them how to encode images and text into the same latent space. Then you simply do a conversion, take a text input, put it into latent space and then extract the image that latent space represents.

topheroo• 1 day ago

I’d argue that talking about “authentic” AI-generated images is oxymoronic.

zzleeper• 1 day ago

Random question, but has there been any improvement in OCR/document understanding in these newer models? Last time I checked (1mo ago) SOTA was still sadly Gemini, unless you wanted to pay $$$ for e.g. Sol

hawtads• 1 day ago

Not sure about OCR specifically, but the newer (past quarter) vision language models all have a lot of post training on detecting garbled text specifically. You can feed some of the old stable diffusion outputs into a modern model and they can figure out pretty reliably if the text gen is mangled. I think the feature is probably used as part of the RL for the image gen to correct for bad text rendering.

geooff_• 1 day ago

No pricing table? No benchmarks? This is just marketingslop.

Image token pricing has been fairly steady while text token prices fall, yet image model release discussion seems to be more focused on how beautiful the women the model generates are versus any sort of substantive discussion.

m3kw9• 1 day ago

Impressive, but their woman face generation always use very similar, too perfect, same prettiness faces, it's very obvious.

dzonga• 1 day ago

once a.i images took over - a lot of things that were based on images died.

going to a fancy restaurant coz of 'you can take good pics' - dead - a.i can recreate that cheaper.

which means for a certain demographic - dating apps are dead too - since those were largely based on swiping photos.

the premium of in-person / small intimate events has gone up. likewise meeting in person, or doing things with a person live.

this also means imperfection has gone up in value (imperfection is a human quality).

hankbond• 1 day ago

the era of wabi-sabi is among us.

BoorishBears• 1 day ago

"the premium of in person" is a string of words that only means something to people in a very specific circle where everyone is constantly trying to tastemake and write in lower case.

Ironically they have extremely limited influence on how the larger world moves: normal people won't meaningfully change their habits around eating good food or meeting up with each other just because AI can generate a picture of either.

jarjoura• 1 day ago

I would attribute any in-person cultural shift to the generation of 20-somethings trapped online during their teenage years from COVID. They are finally old enough, and have enough power to run things.

Mr_Eri_Atlov• 1 day ago

Image generation is the least interesting and most concerning aspect of AI.

I see lots of useful tools made related to translation and code generation, but image generation seems to be primarily used to deceive, harass, or embelish to the point of questioning what the point of a photo is anymore anyway?

kthartic• 1 day ago

I can give one innocent example - I made an app that summarises D&D sessions, including images to depict some of the scenes.

But I understand the jaded perspective. I guess the idea of ‘deepfakes’ didn’t give AI image gen a great reputational start.

lalith_c• 1 day ago

didn’t anthropic accuse them of stealing fable?

flakiness• 1 day ago

It's a bit shocking to see them showing off blatantly disinformation-al examples. See the Japanese manga one. It says "原作監修三浦建太郎" which says its original is written by, and it is supervised the by, the famous author (who died a few years ago so who can he possibly supervise?) and "共同制作白泉社" saying: collaborated by a (famous Japanese) publisher.

I hope these stakeholders have good enough layers.

noodlescb• 1 day ago

It's hilarious how bad image generation is. I just tried this out and I gave it two actual logos, a full design language spec, and three screenshots of the actual application and it spat out three absolutely awful images, none of which had the same logo as the one I sent.

Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge

Related Stories