Not sure why this isn’t a bigger deal —- it seems like this is the first open-source model to beat gpt-image-1 in all respects while also beating Flux Kontext in terms of editing ability. This seems huge.
I've been playing around with it for the past hour. It's really good but from my preliminary testing it definitely falls short of gpt-image-1 (or even Imagen 3/4) where reasonably complex strict prompt adherence is concerned. Scored around ~50% where gpt-image-1 scored ~75%. Couldn't handle the maze, Schrödinger's equation, etc.
Besides style transfer, object additions and removals, text editing, manipulation of human poses, it also supports object detection, semantic segmentation, depth/edge estimation, super-resolution and novel view synthesis (NVS) i.e. synthesizing new perspectives from a base image. It’s quite a smorgasbord!
Early results indicate to me that gpt-image-1 has a bit better sharpness and clarity but I’m honestly not sure if OpenAI doesn’t simply do some basic unsharp mask or something as a post-processing step? I’ve always felt suspicious about that, because the sharpness seems oddly uniform even in out-of-focus areas? And sometimes a bit much, even.
Otherwise, yeah this one looks about as good.
Which is impressive! I thought OpenAI had a lead here from their unique image generation solution that’d last them this year at least.
Oh, and Flux Krea has lasted four days since announcement! In case this one is truly similar in quality to gpt-image-1.
Per 100k image. And it is additionally $0.01 per image. Considering H100 is $1.5 per hour and you can get 1 image per 5s, we are talking about bare-metal cost of ~$0.002 per image + $0.01 license cost.
It's only been a few hours and the demo is constantly erroring out, people need more time to actually play with it before getting excited. Some quantized GGUFs + various comfy workflows will also likely be a big factor for this one since people will want to run it locally but it's pretty large compared to other models. Funnily enough, the main comparison to draw might be between Alibaba and Alibaba. I.e. using Wan 2.2 for image generation has been an extremely popular choice, so most will want to know how big a leap Qwen-Image is from that rather than Flux.
The best time to judge how good a new image model actually is seems to be about a week from launch. That's when enough pieces have fallen into place that people have had a chance to really mess with it and come out with 3rd party pros/cons of the models. Looking hopeful for this one though!
I spun up an H100 on Voltage Park to give it a try in an isolated environment. It's really, really good. The only area where it seems less strong than gpt-image-1 is in generating images of UI (e.g. make me a landing page for Product Hunt in the style of Studio Ghibli), but other than that, I am impressed.
I think the fact that, as far as I understand, it takes 40GB of VRAM to run, is probably dampening some of the enthusiasm.
As an aside, I am not sure why for LLM models the technology to spread among multiple cards is quite mature, while for image models, despite also using GGUFs, this has not been the case. Maybe as image models become bigger there will be more of a push to implement it.
40GB is small IMO: you can run it on a mid-tier Macbook Pro... or the smallest M3 Ultra Mac Studio! You don't need Nvidia if you're doing at-home inference, Nvidia only becomes economical at very high throughput: i.e. dedicated inference companies. Apple Silicon is much more cost effective for single-user for the small-to-medium-sized models. The M3 Ultra is ~roughly on par with a 4090 in terms of memory bandwidth, so it won't be much slower, although it won't match a 5090.
Also for a 20B model, you only really need 20GB of VRAM: FP8 is near-identical to FP16, it's only below FP8 that you start to see dramatic drop-offs in quality. So literally any Mac Studio available for purchase will do, and even a fairly low-end Macbook Pro would work as well. And a 5090 should be able to handle it with room to spare as well.
Ah, you're right: it doesn't have dedicated FP8 cores, so you'd get significantly worse performance (a quick Google search implies 5x worse). Although you could still run the model, just slowly.
Any M3 Ultra Mac Studio, or midrange-or-better Macbook Pro, would handle FP16 with no issues though. A 5090 would handle FP8 like a champ and a 4090 could probably squeeze it in as well, although it'd be tight.
> I think the fact that, as far as I understand, it takes 40GB of VRAM to run, is probably dampening some of the enthusiasm.
40 GB of VRAM? So two GPU with 24 GB each? That's pretty reasonable compared to the kind of machine to run the latest Qwen coder (which btw are close to SOTA: they do also beat proprietary models on several benchmarks).
A 3090 + 2xTitanXP? technically i have 48, but i don't think you can "split it" over multiple cards. At least with Flux, it would OOM the Titans and allocate the full 3090
With the notable exception of gpt-image-1, discussion about AI image generation has become much less popular. I suspect it's a function of a) AI discourse being dominated by AI agents/vibe coding and b) the increasing social stigma of AI image generation.
Flux Kontext was a gamechanger release for image editing and it can do some absurd things, but it's still relatively unknown. Qwen-Image, with its more permissive license, could lead to much more innovation once the editing model is released.
There's no social stigma to using AI image generation.
There is what's probably better described as a bullying campaign. People tried the same thing when synthesizers and cameras were invented. But nobody takes it seriously unless you're already in the angry person fandom.
In practice AI image generation is ubiquitous at this point. AI image editing is also built into all major phones.
Useless AI art (which is almost all of it) is not like the camera or the synthesizer, it's closer to when 50-60yo moms were sharing Minion memes on facebook: cringe and tasteless. It getting better won't make it more accepted, it will simply make people suspect of actual art until no one really gives a chance to any of it.
Your argument might actually be suggesting that you don't like art in general more than that there is a stigma against AI. If there is no value in artisanal art that differentiates it from AI-produced works and therefore both will be discarded as the quality converges, what was supposed to be the value in art to start with?
There absolutely is - everytime someone uses an AI image in a presentation slide, or in an article to illustrate the point, everybody just rolls their eyes - in my opinion a stock photo or even nothing is preferable to a low effort AI image.
Responding to myself, as I realized that my post above feels too dismissive. Being a long time privacy advocate for non-tech-adjacent people, I'm perfectly aware about my bubble and biases. For any normal person, anything I say about digital privacy sounds absolutely abstract and detached from real life, where convenience and low effort dominates everything else. Even in 2025 with all political shenanigans, they just fail to see the link and how it applies to their life. AI imagegen is the same from my observations, most concerns are contained in a tiny bubble of perpetually online people. Not even all artists share the loud opinions (for reference, I used to manage a couple hundred artists), especially not VFX and 3D folks. And that tiny bubble only really exists in the anglosphere - you'll see a completely different picture in other cultural bubbles. There's absolutely no stigma of any kind outside of it.
Social stigma? Only if you listen to mentally ill Twitter users.
It's more that the novelty just wore off. Mainstream image generation in online services is "good enough" for most casual users - and power users are few, and already knee deep in custom workflows. They aren't about to switch to the shiny new thing unless they see a lot of benefits to it.
Considering they have not released their image, editor weights, I’m not sure how you could make a conclusion that it is better than Flux Kontext aside from the graphs they put out.
But, obviously you wouldn’t do that. Right? Did you look at the scaling on their graphs?
Good release! I've added it to the GenAI Showdown site. Overall a pretty good model scoring around 40% - and definitely represents SOTA for something that could be reasonably hosted on consumer GPU hardware (even more so when its quantized).
That being said, it still lags pretty far behind OpenAI's gpt-image-1 strictly in terms of prompt adherence for txt2img prompting. However as has already been mentioned elsewhere in the thread, this model can do a lot more around editing, etc.
Even though I didn't see a significant improvement over Imagen3 in adherence, I agree. Initially the page was just getting a bit crowded but now that I've added a "Show/Hide Models" toggle I'll go ahead and make that change.
In their own first example of English text rendering, it's mistakenly rendered "The silent patient" as "The silent Patient", "The night circus" as "The night Circus", and miskerned "When stars are scattered" as "When stars are sca t t e r e d".
The example further down has "down" not "dawn" in the poem.
For these to be their hero image examples, they're fairly poor; I know it's a significant improvement vs. many of the other current offerings, but it's clear the bar is still being set very low.
The fact that it doesn’t change the images like 4o image gen is incredible. Often when I try to tweak someone’s clothing using 4o, it also tweaks their face. This only seems to apply those recognizable AI artifacts to only the elements needing to be edited.
This may be obvious to people who do this regularly, but what kind of machine is required to run this? I downloaded & tried it on my Linux machine that has a 16GB GPU and 64GB of RAM. This machine can run SD easily. But Qwen-image ran out of space both when I tried it on the GPU and on the CPU, so that's obviously not enough. But am I off by a factor of two? An order of magnitude? Do I need some crazy hardware?
> This may be obvious to people who do this regularly
This is not that obvious. Calculating VRAM usage for VLMs/LLMs is something of an arcane art. There are about 10 calculators online you can use and none of them work. Quantization, KV caching, activation, layers, etc all play a role. It's annoying.
But anyway, for this model, you need 40+ GB of VRAM. System RAM isn't going to cut it unless it's unified RAM on Apple Silicon, and even then, memory bandwidth is shot, so inference is much much slower than GPU/TPU.
Also I think you need a 40GB "card", not just 40GB of vram. I wrote about this upthread, you're probably going to need one card, I'd be surprised if you could chain several GPUs together.
Oh right, I forgot some diffusion models can't offload / split layers. I don't use vision generation models much at all - was just going off LLM work. Apologies for the potential misinformation.
Nah, that won’t gain you much (if anything?) over just doing the layer swaps on RAM. You can put the text encoder on the second card but you can also just put it in your RAM without much for negatives.
I believe it's roughly the same size as the model files. If you look in the transformers folder you can see there are around 9 5gb files, so I would expect you need ~45gb vram on your GPU. Usually quantized versions of models are eventually released/created that can run on much less vram but with some quality loss.
I've been bugging them about this for a while. There are repos that contain multiple model weights in a single repo which means adding up the file sizes won't work universally, but I'd still find it useful to have a "repo size" indicator somewhere.
> I think the fact that, as far as I understand, it takes 40GB of VRAM to run, is probably dampening some of the enthusiasm.
For PCs I take it one that has two PCIe 4.0 x16 or more recent slots? As in: quite some consumers motherboards. You then put two GPU with 24 GB of VRAM each.
A friend runs this (don't know if the tried this Qwen-Image yet): it's not an "out of this world" machine.
Does anyone know how they actually trained text rendering into these models?
To me they all seem to suffer from the same artifacts, that the text looks sort of unnatural and doesn't have the correct shadows/reflections as the rest of the image. This applies to all the models I have tried, from OpenAI to Flux. Presumably they are all using the same trick?
It's on page 14 of the technical report. They generate synthetic data by putting text on top of an image, apparently without taking the original lighting into account. So that's the look the model reproduces. Garbage in, garbage out.
Maybe in the future someone will come up with a method for putting realistic text into images so that they can generate data to train a model for putting realistic text into images.
If you think diffusing legible, precise text from pure noise is garbage then wtf are you doing here. The arrogance of the it crowd can be staggering at times
I love that this is the only thing the community wants to know at every announce of a new model, but no organization wants to face the crude reality of human nature.
That, and the weird prudishness of most american people and companies.
I love this attitude of Americans, haha. They ignore that their representative companies only give them completely black-box APIs to use, yet they nitpick these open-weight models. Their own country is controlled by AIPAC, immorally complicit in genocide, but on the other hand they condescendingly criticize China. Haha, enjoy your last glorious moments as a hegemony.
> In this case, the paper is less than one-tenth of the entire image, and the paragraph of text is relatively long, but the model still accurately generates the text on the paper.
Nope. The text includes the line "That dawn will bloom" but the render reads "That down will bloom", which is meaningless.
the value is: the absence of text where you expect it, and the presence of garbled text, are dead giveaways of AI generation. i'm not sure why you are being downvoted, compositing text seems like a legitimate alternative.
it seems like the value is that you don't need another tool to composite the text. especially for users who aren't aware of figma/photoshop nor how to use them (many many many people)
And if you want the text to faithfully follow the surface of the object (ex tattoos) I don't think the post AI gen manual editing approach is going to be so straightforward.
I’m interested to see what this model can do, but also kinda annoyed at the use of a Studio Ghibli style image as one of the first examples. Miyazaki has said over and over that he hates AI image generation. Is it really so much to ask that people not deliberately train LoRAs and finetunes specifically on his work and use them in official documentation?
It reminds me of how CivitAI is full of “sexy Emma Watson” LoRAs, presumably because she very notably has said she doesn’t want to be portrayed in ways that objectify her body. There’s a really rotten vein of “anti-consent” pulsing through this community, where people deliberately seek out people who have asked to be left out of this and go “Oh yeah? Well there’s nothing you can do to stop us, here’s several terabytes of exactly what you didn’t want to happen”.
> Miyazaki has said over and over that he hates AI image generation
No he has not. He was talking about an AI model that was shown off for crudely animating 3D people in 2016, in a way that he found creepy. If you watch the actual video, you can see the examples that likely set him off here[0].
It's all too much of cringe. AI creativity space is chock full of cringy cargocult parody of "no such things as bad publicity" strategy. Things on the Internet is reposted to death so what's wrong if we use them what even is copyright. Everybody hates AI generated images sure that's how you get the word out. Pornography drives adoption so let them have some it should work.
Those behaviors might appear correct in an extremely superficial sense, but it is as if they prompted themselves for "man eating cookies" and ended up with what is akin to early Will Smith pasta gifs. Whatever they're doing and assuming it's cookies held in hands, they're not eating them.
I mean, did you really expect anything more from the internet? Maybe I'm wrong, but hentai, erotic roleplay, and nudify applications seem to still represent a massive portion of AI use cases. At least in the case of ero RP, perhaps the exploitation of people for pornography might be lessened....
I get that if you can imagine something, it exists, and also there is porn of it.
What disappoints me is how aligned the whole community is with its worst exponents. That someone went “Heh heh, I’m gonna spend hours of my day and hundreds/thousands of dollars in compute just to make Miyazaki sad.” and then influencers in the AI art space saw this happen and went “Hell yeah let’s go” and promoted the shit out of it making it one of the few finetunes to actually get used by normies in the mainstream, and then leaders in this field like the Qwen team went “Yeah sure let’s ride the wave” and made a Studio Ghibli style image their first example.
I get that there was no way to physically stop a Studio Ghibli LoRA from existing. I still think the community’s gleeful reaction to it has been gross.
People are downvoting you but it's true. Ghibli is just the highest profile studio that creates work in that general style. Arguably most of the highest quality examples of that style are their work. However they're far from the only practitioners.
They have a much better and cleaner dataset than Stable Diffusion & others, so I’d expect it to be better with some kinds of images (photos in particular)
as long as you don't consider the part of the model which understands text as part of the model, and as long as you don't consider copyrighted text content copyrighted :)
Not sure why this isn’t a bigger deal —- it seems like this is the first open-source model to beat gpt-image-1 in all respects while also beating Flux Kontext in terms of editing ability. This seems huge.
I've been playing around with it for the past hour. It's really good but from my preliminary testing it definitely falls short of gpt-image-1 (or even Imagen 3/4) where reasonably complex strict prompt adherence is concerned. Scored around ~50% where gpt-image-1 scored ~75%. Couldn't handle the maze, Schrödinger's equation, etc.
https://genai-showdown.specr.net
Great work, thanks.
Midjourneys images are the only ones which don’t make me uncomfortable (most of the time), hopefully they can fix their prompt adherence.
interesting to see how many still can't handle the nine pointed star correctly
Fantastic comparisons! Great to see the limitations of the latest models.
Prompt idea: "A person holding a wooden Penrose triangle." Only GPT-4o image generation is able to make Penrose triangles, as far as I can tell.
Is it fair to call the OpenAI octopus “real”?
Those are the wrong images - the CDN was caching older media - I've since purged it so the right ones should show up now. Thanks for the call out!
I think it does way more than gpt-image-1 too?
Besides style transfer, object additions and removals, text editing, manipulation of human poses, it also supports object detection, semantic segmentation, depth/edge estimation, super-resolution and novel view synthesis (NVS) i.e. synthesizing new perspectives from a base image. It’s quite a smorgasbord!
Early results indicate to me that gpt-image-1 has a bit better sharpness and clarity but I’m honestly not sure if OpenAI doesn’t simply do some basic unsharp mask or something as a post-processing step? I’ve always felt suspicious about that, because the sharpness seems oddly uniform even in out-of-focus areas? And sometimes a bit much, even.
Otherwise, yeah this one looks about as good.
Which is impressive! I thought OpenAI had a lead here from their unique image generation solution that’d last them this year at least.
Oh, and Flux Krea has lasted four days since announcement! In case this one is truly similar in quality to gpt-image-1.
Not to mention, flux models are for non-commercial use only.
the license for flux models is $1,000/mo, hardly an obstacle to any serious commercial usage
Per 100k image. And it is additionally $0.01 per image. Considering H100 is $1.5 per hour and you can get 1 image per 5s, we are talking about bare-metal cost of ~$0.002 per image + $0.01 license cost.
The pricing seems reasonable for a SOTA class model that needs to be commercially viable or it dies.
It's not clear from their page but the editing model is not released yet: https://github.com/QwenLM/Qwen-Image/issues/3#issuecomment-3...
It's only been a few hours and the demo is constantly erroring out, people need more time to actually play with it before getting excited. Some quantized GGUFs + various comfy workflows will also likely be a big factor for this one since people will want to run it locally but it's pretty large compared to other models. Funnily enough, the main comparison to draw might be between Alibaba and Alibaba. I.e. using Wan 2.2 for image generation has been an extremely popular choice, so most will want to know how big a leap Qwen-Image is from that rather than Flux.
The best time to judge how good a new image model actually is seems to be about a week from launch. That's when enough pieces have fallen into place that people have had a chance to really mess with it and come out with 3rd party pros/cons of the models. Looking hopeful for this one though!
I spun up an H100 on Voltage Park to give it a try in an isolated environment. It's really, really good. The only area where it seems less strong than gpt-image-1 is in generating images of UI (e.g. make me a landing page for Product Hunt in the style of Studio Ghibli), but other than that, I am impressed.
I think the fact that, as far as I understand, it takes 40GB of VRAM to run, is probably dampening some of the enthusiasm.
As an aside, I am not sure why for LLM models the technology to spread among multiple cards is quite mature, while for image models, despite also using GGUFs, this has not been the case. Maybe as image models become bigger there will be more of a push to implement it.
40GB is small IMO: you can run it on a mid-tier Macbook Pro... or the smallest M3 Ultra Mac Studio! You don't need Nvidia if you're doing at-home inference, Nvidia only becomes economical at very high throughput: i.e. dedicated inference companies. Apple Silicon is much more cost effective for single-user for the small-to-medium-sized models. The M3 Ultra is ~roughly on par with a 4090 in terms of memory bandwidth, so it won't be much slower, although it won't match a 5090.
Also for a 20B model, you only really need 20GB of VRAM: FP8 is near-identical to FP16, it's only below FP8 that you start to see dramatic drop-offs in quality. So literally any Mac Studio available for purchase will do, and even a fairly low-end Macbook Pro would work as well. And a 5090 should be able to handle it with room to spare as well.
If you want to wait 20 minutes for one image you can certainly run it on a macbook pro.
The quality doesn't have to get much higher for that to be a great deal. For humans the wait time is typically measured in days.
Memory bandwidth is only relevant for comparing LLM performance. For image generation, the limiting factor is compute, and Apple sucks with it.
Does M3 Ultra or later have hardware FP8 support on the CPU cores?
Ah, you're right: it doesn't have dedicated FP8 cores, so you'd get significantly worse performance (a quick Google search implies 5x worse). Although you could still run the model, just slowly.
Any M3 Ultra Mac Studio, or midrange-or-better Macbook Pro, would handle FP16 with no issues though. A 5090 would handle FP8 like a champ and a 4090 could probably squeeze it in as well, although it'd be tight.
If 40GB you can lightly quantize and fit it on a 5090.
Which very few people have, comparatively.
Training it will also be out of reach for most. I’m sure I’ll be able to handle it on my own 5090 at some point but it’ll be slow going.
> I think the fact that, as far as I understand, it takes 40GB of VRAM to run, is probably dampening some of the enthusiasm.
40 GB of VRAM? So two GPU with 24 GB each? That's pretty reasonable compared to the kind of machine to run the latest Qwen coder (which btw are close to SOTA: they do also beat proprietary models on several benchmarks).
A 3090 + 2xTitanXP? technically i have 48, but i don't think you can "split it" over multiple cards. At least with Flux, it would OOM the Titans and allocate the full 3090
You can’t split image models over 2 GPUs like you can LLMs.
They also released an inference server for their models. Wan and qwen-image can be split without problems. https://github.com/modelscope/DiffSynth-Engine
Have you tried the image editing?
With the notable exception of gpt-image-1, discussion about AI image generation has become much less popular. I suspect it's a function of a) AI discourse being dominated by AI agents/vibe coding and b) the increasing social stigma of AI image generation.
Flux Kontext was a gamechanger release for image editing and it can do some absurd things, but it's still relatively unknown. Qwen-Image, with its more permissive license, could lead to much more innovation once the editing model is released.
There's no social stigma to using AI image generation.
There is what's probably better described as a bullying campaign. People tried the same thing when synthesizers and cameras were invented. But nobody takes it seriously unless you're already in the angry person fandom.
In practice AI image generation is ubiquitous at this point. AI image editing is also built into all major phones.
Useless AI art (which is almost all of it) is not like the camera or the synthesizer, it's closer to when 50-60yo moms were sharing Minion memes on facebook: cringe and tasteless. It getting better won't make it more accepted, it will simply make people suspect of actual art until no one really gives a chance to any of it.
Your argument might actually be suggesting that you don't like art in general more than that there is a stigma against AI. If there is no value in artisanal art that differentiates it from AI-produced works and therefore both will be discarded as the quality converges, what was supposed to be the value in art to start with?
There absolutely is - everytime someone uses an AI image in a presentation slide, or in an article to illustrate the point, everybody just rolls their eyes - in my opinion a stock photo or even nothing is preferable to a low effort AI image.
Who is everybody? How do you know? What is your personal bubble? Could it be you're presenting your opinion for the thing that commonly happens?
Responding to myself, as I realized that my post above feels too dismissive. Being a long time privacy advocate for non-tech-adjacent people, I'm perfectly aware about my bubble and biases. For any normal person, anything I say about digital privacy sounds absolutely abstract and detached from real life, where convenience and low effort dominates everything else. Even in 2025 with all political shenanigans, they just fail to see the link and how it applies to their life. AI imagegen is the same from my observations, most concerns are contained in a tiny bubble of perpetually online people. Not even all artists share the loud opinions (for reference, I used to manage a couple hundred artists), especially not VFX and 3D folks. And that tiny bubble only really exists in the anglosphere - you'll see a completely different picture in other cultural bubbles. There's absolutely no stigma of any kind outside of it.
Social stigma? Only if you listen to mentally ill Twitter users.
It's more that the novelty just wore off. Mainstream image generation in online services is "good enough" for most casual users - and power users are few, and already knee deep in custom workflows. They aren't about to switch to the shiny new thing unless they see a lot of benefits to it.
gpt-image-1 is the League of Legends of image generation. It is a tool in front of like 30 million DAUs...
Slightly hyperbolic, gpt-image-1 is better on at least a couple of the text metrics.
Considering they have not released their image, editor weights, I’m not sure how you could make a conclusion that it is better than Flux Kontext aside from the graphs they put out.
But, obviously you wouldn’t do that. Right? Did you look at the scaling on their graphs?
how can it beat gpt-image-1 if there is no image editor?
Do try it. The image quality and diversity is pretty shocking and not in a good way.
Good release! I've added it to the GenAI Showdown site. Overall a pretty good model scoring around 40% - and definitely represents SOTA for something that could be reasonably hosted on consumer GPU hardware (even more so when its quantized).
That being said, it still lags pretty far behind OpenAI's gpt-image-1 strictly in terms of prompt adherence for txt2img prompting. However as has already been mentioned elsewhere in the thread, this model can do a lot more around editing, etc.
https://genai-showdown.specr.net
Side remark: I don't think it's appropriate to mix Imagen 3 and 4. Those are two different models.
Even though I didn't see a significant improvement over Imagen3 in adherence, I agree. Initially the page was just getting a bit crowded but now that I've added a "Show/Hide Models" toggle I'll go ahead and make that change.
In their own first example of English text rendering, it's mistakenly rendered "The silent patient" as "The silent Patient", "The night circus" as "The night Circus", and miskerned "When stars are scattered" as "When stars are sca t t e r e d".
The example further down has "down" not "dawn" in the poem.
For these to be their hero image examples, they're fairly poor; I know it's a significant improvement vs. many of the other current offerings, but it's clear the bar is still being set very low.
The fact that it doesn’t change the images like 4o image gen is incredible. Often when I try to tweak someone’s clothing using 4o, it also tweaks their face. This only seems to apply those recognizable AI artifacts to only the elements needing to be edited.
That's why Flux Kontext was such a huge deal - it gave you the power of img2img inpainting without needing to manually mask the content.
https://mordenstar.com/blog/edits-with-kontext
Seems strange to not include the prompts themselves, if people are curious in trying to replicate it themselves.
You can select the area you want edited on 4o, and it’ll keep the rest unchanged
gpt doesn't respect masks
Correct. Have tried this without much success despite OpenAI's claims.
This may be obvious to people who do this regularly, but what kind of machine is required to run this? I downloaded & tried it on my Linux machine that has a 16GB GPU and 64GB of RAM. This machine can run SD easily. But Qwen-image ran out of space both when I tried it on the GPU and on the CPU, so that's obviously not enough. But am I off by a factor of two? An order of magnitude? Do I need some crazy hardware?
> This may be obvious to people who do this regularly
This is not that obvious. Calculating VRAM usage for VLMs/LLMs is something of an arcane art. There are about 10 calculators online you can use and none of them work. Quantization, KV caching, activation, layers, etc all play a role. It's annoying.
But anyway, for this model, you need 40+ GB of VRAM. System RAM isn't going to cut it unless it's unified RAM on Apple Silicon, and even then, memory bandwidth is shot, so inference is much much slower than GPU/TPU.
Also I think you need a 40GB "card", not just 40GB of vram. I wrote about this upthread, you're probably going to need one card, I'd be surprised if you could chain several GPUs together.
Oh right, I forgot some diffusion models can't offload / split layers. I don't use vision generation models much at all - was just going off LLM work. Apologies for the potential misinformation.
Not sure what you mean or new to llms, but two RTX 3090 will work for this, and even lower-end cards will (RTX3060) once it's GGUF'd
This isn't a transformer, it's a diffusion model. You can't split diffusion models across compute nodes.
do you mean https://github.com/pollockjj/ComfyUI-MultiGPU? One GPU would do the computation, but others could pool in for VRAM expansion, right? (I've not used this node)
Nah, that won’t gain you much (if anything?) over just doing the layer swaps on RAM. You can put the text encoder on the second card but you can also just put it in your RAM without much for negatives.
will the new AMD AI CPUs work? like an AI HX 395 or the slower 370? I'm stuck on an A2000 w/16GB of VRAM and wondering what's a worthwhile upgrade.
It may fit but image generation on anything but Nvidia is so slow it won’t be worth it.
I believe it's roughly the same size as the model files. If you look in the transformers folder you can see there are around 9 5gb files, so I would expect you need ~45gb vram on your GPU. Usually quantized versions of models are eventually released/created that can run on much less vram but with some quality loss.
Model size = file for fp8, so if this was released at fp16 then 40-ish, if it's quantized to fp4 then 10ish
Why doesn't huggingface list the aggregate model size?
I've been bugging them about this for a while. There are repos that contain multiple model weights in a single repo which means adding up the file sizes won't work universally, but I'd still find it useful to have a "repo size" indicator somewhere.
I ended up building my own tool for that: https://tools.simonwillison.net/huggingface-storage
I've been wondering this for literally years now...
Huggingface is just a git hosting service, like github. You can add up the sizes of all the files in the directory yourself
That’s what we have computers for though - to compute.
You're probably going to have to wait a couple of days for 4 bit quantized versions to pop up. It's 20B parameters.
update: doesn't work well. this approach seems to be recommended: https://github.com/QwenLM/Qwen-Image/pull/6/files
Qwen-Image requires at least 24GB VRAM for the full model, but you can run the 4-bit quantized version with ~8GB VRAM using libraries like AutoGPTQ.
16GiB RAM with 8-bit quantization.
This is a slightly scaled up SD3 Large model (38 layers -> 60 layers).
For prod inference, 1xH100 is working well.
two p40 cards together will run this for under $300
> I think the fact that, as far as I understand, it takes 40GB of VRAM to run, is probably dampening some of the enthusiasm.
For PCs I take it one that has two PCIe 4.0 x16 or more recent slots? As in: quite some consumers motherboards. You then put two GPU with 24 GB of VRAM each.
A friend runs this (don't know if the tried this Qwen-Image yet): it's not an "out of this world" machine.
maybe not "out of this world" but still not cheap. probably $4,000 with 3090s. pretty big chunk of change for some ai pictures.
You can’t split diffusion models like that.
I just tested it out, very impressive results. I wonder what the Queen team did behind the scenes to make this work so well.
https://chat.qwen.ai/
(Select "Image Generation" and be sure to use the Qwen3-235B model - also tried selecting "Coder" but it errors out.)
None of the image model could handle showing time like generate a clock showing 3:15 pm.
Yeah, clock times are a notorious challenge for LLMs due to training data almost always showing aesthetically appealing times like 10-and-2.
An entire thread on this subject previously unfolded on HN but I can't find it at this time!
Does anyone know how they actually trained text rendering into these models?
To me they all seem to suffer from the same artifacts, that the text looks sort of unnatural and doesn't have the correct shadows/reflections as the rest of the image. This applies to all the models I have tried, from OpenAI to Flux. Presumably they are all using the same trick?
It's on page 14 of the technical report. They generate synthetic data by putting text on top of an image, apparently without taking the original lighting into account. So that's the look the model reproduces. Garbage in, garbage out.
Maybe in the future someone will come up with a method for putting realistic text into images so that they can generate data to train a model for putting realistic text into images.
Wouldn't it make sense to use rendered images for that?
i'm not sure if that's such garbage as you suggest, surely it is helpful for generalization yes? kind of the point of self-supervised models
If you think diffusing legible, precise text from pure noise is garbage then wtf are you doing here. The arrogance of the it crowd can be staggering at times
They're referring to the training data being garbage, not the diffusion process.
Insane how many good Chinese open source models they've been releasing. This really gives me hope
Taking a concrete lead in LLM-world would be a big national win for China.
I have the impression this might be a strategy to help boost the AI bubble. Big tech capex rn is too big to fail
Checkout section 3.2 Data Filtering:
https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/Q...
It's also kind of interesting that no other languages than English and Chinese are named or shown...
Jaw dropping. Because text rendering isn't easy even with regular programming SDKs etc.
Anyone thinking otherwise hasn't attempted implementing it or haven't thought about it in depth.
A beast. Supposedly beats GPT-4o in image generation and Flux Kontext in image editing.
If it’s as good as they say, one less reason for that ChatGPT sub..
How censored is it?
I love that this is the only thing the community wants to know at every announce of a new model, but no organization wants to face the crude reality of human nature.
That, and the weird prudishness of most american people and companies.
I love this attitude of Americans, haha. They ignore that their representative companies only give them completely black-box APIs to use, yet they nitpick these open-weight models. Their own country is controlled by AIPAC, immorally complicit in genocide, but on the other hand they condescendingly criticize China. Haha, enjoy your last glorious moments as a hegemony.
Short canva.
> In this case, the paper is less than one-tenth of the entire image, and the paragraph of text is relatively long, but the model still accurately generates the text on the paper.
Nope. The text includes the line "That dawn will bloom" but the render reads "That down will bloom", which is meaningless.
The text rendering is impressive, but I don't understand the value — wouldn't it be easier to add any text that you like in Figma?
If you mass-publish chatgpt generated books on amazon it might be pretty useful.
the value is: the absence of text where you expect it, and the presence of garbled text, are dead giveaways of AI generation. i'm not sure why you are being downvoted, compositing text seems like a legitimate alternative.
it seems like the value is that you don't need another tool to composite the text. especially for users who aren't aware of figma/photoshop nor how to use them (many many many people)
And if you want the text to faithfully follow the surface of the object (ex tattoos) I don't think the post AI gen manual editing approach is going to be so straightforward.
I’m interested to see what this model can do, but also kinda annoyed at the use of a Studio Ghibli style image as one of the first examples. Miyazaki has said over and over that he hates AI image generation. Is it really so much to ask that people not deliberately train LoRAs and finetunes specifically on his work and use them in official documentation?
It reminds me of how CivitAI is full of “sexy Emma Watson” LoRAs, presumably because she very notably has said she doesn’t want to be portrayed in ways that objectify her body. There’s a really rotten vein of “anti-consent” pulsing through this community, where people deliberately seek out people who have asked to be left out of this and go “Oh yeah? Well there’s nothing you can do to stop us, here’s several terabytes of exactly what you didn’t want to happen”.
> Miyazaki has said over and over that he hates AI image generation
No he has not. He was talking about an AI model that was shown off for crudely animating 3D people in 2016, in a way that he found creepy. If you watch the actual video, you can see the examples that likely set him off here[0].
[0] https://youtu.be/ngZ0K3lWKRc&t=7
It's all too much of cringe. AI creativity space is chock full of cringy cargocult parody of "no such things as bad publicity" strategy. Things on the Internet is reposted to death so what's wrong if we use them what even is copyright. Everybody hates AI generated images sure that's how you get the word out. Pornography drives adoption so let them have some it should work.
Those behaviors might appear correct in an extremely superficial sense, but it is as if they prompted themselves for "man eating cookies" and ended up with what is akin to early Will Smith pasta gifs. Whatever they're doing and assuming it's cookies held in hands, they're not eating them.
It's just really distasteful and unoriginal. I cringe inside whenever I see a Ghibli-style profile picture. Have some originality for god sake.
Leading by example by not condoning copying artists' styles would be a simple polite gesture.
I mean, did you really expect anything more from the internet? Maybe I'm wrong, but hentai, erotic roleplay, and nudify applications seem to still represent a massive portion of AI use cases. At least in the case of ero RP, perhaps the exploitation of people for pornography might be lessened....
I get that if you can imagine something, it exists, and also there is porn of it.
What disappoints me is how aligned the whole community is with its worst exponents. That someone went “Heh heh, I’m gonna spend hours of my day and hundreds/thousands of dollars in compute just to make Miyazaki sad.” and then influencers in the AI art space saw this happen and went “Hell yeah let’s go” and promoted the shit out of it making it one of the few finetunes to actually get used by normies in the mainstream, and then leaders in this field like the Qwen team went “Yeah sure let’s ride the wave” and made a Studio Ghibli style image their first example.
I get that there was no way to physically stop a Studio Ghibli LoRA from existing. I still think the community’s gleeful reaction to it has been gross.
Whatever. "Studio Ghibli style" is so loose of a definition to begin with. You can't own a "style" anyway. Tough cookies.
People are downvoting you but it's true. Ghibli is just the highest profile studio that creates work in that general style. Arguably most of the highest quality examples of that style are their work. However they're far from the only practitioners.
Seems a bit drastic to compare Ghibli style transfer to revenge porn, but you do you I guess.
It’s the anti-consent thing that ties them together. The idea of “You asked us to leave you alone, which is why we’re targeting you.”
Why are you talking about revenge porn here?
Welcome to the internet, which is for porn (and cat pictures).
is it an official site? https://qwen-image.ai
What lowest graphic card can support this self hosted with a reasonable output !
Probably a 4080 when the nunchaku quants drop.
Wow, the text/writing is amazing! Also the editing in general, but the text really stands out
It will take years for people to use these but Adobe is not alone.
Adobe has never been alone. Photoshop’s AI stuff is consistently behind OSS models and workflows. It’s just way more convenient
I think Adobe is also very careful with copyrighted content not being a part of their models, which inherently makes them of lower quality.
They have a much better and cleaner dataset than Stable Diffusion & others, so I’d expect it to be better with some kinds of images (photos in particular)
as long as you don't consider the part of the model which understands text as part of the model, and as long as you don't consider copyrighted text content copyrighted :)
Team Qwen: Please stop ripping off Studio Ghibli to demo your product.
That entire banner is pure copyright infringement.