71–80 of 127 posts
The Original model shows the FRONT, the speed version shows the BACK of the corvette. It's a completely different picture. This is not similar but strikingly different.
I just cancelled my Midjourney subscription, it feels like it's fallen too far behind for the stuff I'd like to do. Spent a lot of time considering using Replicate as well as Ideogram.
never used midjourney because it had that signature look and bad with hands, feet, letters crazy not even a year has past since Emad's downfall a local open source and superior model drops which just shows how little moat these companies have and are just lighting cash on fire which we benefit from
> which just shows how little moat these companies have
Flux was developed by the same people that made Stable Diffusion.
Text to image models feels inefficient to me. I wonder if it would be possible and better to do it in separate steps, like text to scene graph, scene graph to semantically segmented image, segmented image to final image. That way each step could be trained separately and be modular, and the image would be easier to edit instead of completely replace it with the output of a new prompt. That way it should be much easie…
If you can train a neural network that goes from a to b and a network that goes from b to c, you can usually replace that combination with a simpler network that goes from a to c directly.
This makes sense, as there might be information in a that we lose by a conversion to b. A single neural network will ensure that all relevant information from a that we need to generate c will be passed to the upper layers.
Earlier quoted context omitted.
The issue with this is there's a false assumption that an image is a collection of objects. It's not (necessarily). I want a picture of frozen cyan peach fuzz.
https://imgur.com/ayAWSKr Prompt: frozen cyan peach fuzz, with default settings on a first generation SD model. People _seriously_ do not understand how good these tools have been for nearly two years already.
Earlier quoted context omitted.
Isn't this essemtially the approach to image recognition etc. that failed for ages until we brute forced it with bigger and deeper matrices? It seems sensible to extract features and reason about things the way a human would, but it turns out its easier to scale pattern matching purely done by computer.
This is Sutton's Bitter Lesson : https://www.cs.utexas.edu/~eunsol/courses/data/bitter_lesson...
My point being, optimization or splitting up int subs, before handing over the problem to the machine, makes sense.
Non-commercial is not open-source, because if the original copyright holder stops maintaining it, nobody else can continue (or has to work like a slave for free). Open-source is about what happens if the original author stops working on it. Open-source gives everyone the license to continue developing it, which obviously means also the ability to get paid. Don't call it open-source if this aspect is missing. Only the…
Doesn’t open source mean the source is viewable/inspectable? I don’t know any closed source apps that let you view the source.
According to the OSI definition, you also need a right to modify the source and/or distribute patches.
> I don’t know any closed source apps that let you view the source.
A lot of them do, especially in the open-core space. THe model is called source-available.
If you're selling to enterprises and not gamers, that model makes sense. What stops large enterprises from pirating software is their own lawyers, not DRM.
This is why you can put a lot of strange provisions into enterprise software licenses, even if you have little to no way to enforce these provisions on a purely technical level.
Text to image models feels inefficient to me. I wonder if it would be possible and better to do it in separate steps, like text to scene graph, scene graph to semantically segmented image, segmented image to final image. That way each step could be trained separately and be modular, and the image would be easier to edit instead of completely replace it with the output of a new prompt. That way it should be much easie…
Isn't this essemtially the approach to image recognition etc. that failed for ages until we brute forced it with bigger and deeper matrices? It seems sensible to extract features and reason about things the way a human would, but it turns out its easier to scale pattern matching purely done by computer.
With image generation on the other hand, which starts from a handful of words, we can first do some text processing into categories, such as objects vs people, color vs brightness, environment vs main object, etc.
If you want to play with FLUX.schnell easily, type the prompt into a Pollinations URL: https://pollinations.ai/p/a_donkey_holding_a_sign_with_flux_... https://pollinations.ai/p/a_donkey_holding_a_sign_with_flux_... https://pollinations.ai/p/Minimalist%20and%20conceptual%20ar... It's incredible how fast it is. We generate 8000 images every 30 minutes for our users using only three L40S GPUs. Disclaimer: I'm behind Pol…
The "only" word sounds quite expensive for most of us.
Non-commercial is not open-source, because if the original copyright holder stops maintaining it, nobody else can continue (or has to work like a slave for free). Open-source is about what happens if the original author stops working on it. Open-source gives everyone the license to continue developing it, which obviously means also the ability to get paid. Don't call it open-source if this aspect is missing. Only the…
Doesn’t open source mean the source is viewable/inspectable? I don’t know any closed source apps that let you view the source.