Cards against Cogniwerk - The tool we don’t deserve, but the tool we need.

As a startup it is crucial for us to fight against our flaws. To fall and stand up again repeatedly. And for a better and faster way of falling we decide to face our fears and ask people from the creative industry to criticize our service and give opinions on what to do better. This is about improving with every feedback. So thank you for participating and enjoy reading. If you want to be part of this series please reach out to us by sending your own words to cardsagainst@cogniwerk.ai

The tool we don’t deserve, but the tool we need.

by Benjamin, founder and creative technologist at Cogniwerk

It’s monday morning, today is testing day and we will test the newest features. We will have some new features which will improve the interface and make it better to understand how it works. And at the same I read in my social media profile about how Midjourney will be incorporated into X and Grok bringing it even more to the masses. 

Yes, we’re fighting against a Goliath, but what are we fighting for? And for whom?

If people think of AI, they think of ChatGPT and Midjourney. They do not think of Comfy, IPadapter, Wave2Lip or even Langchain. These things are of no interest, they are deemed too special, too nerdy - and yet they are used by millions but not visible because they are below interfaces, which are so clean and easy, it makes look Adobe products like some behemoth of past ages. And creatives use these easy tools. Sure they want control, but they do not want to really invest more time then it takes for a google query. The average creative person who is also interested in AI does not really care about the wonders of data and latent spaces. For them AI is the promise of making the formerly hard stuff easy. Not to have the rest of day off and engage in civil services, spending time with their family or working for NGOs. No, to tackle more projects or enhance “the quality”. 

And this “quality” does not lie in diverse visualizations but eventually… stock pictures. Tools are rated by their output to produce “stunning” images, which are actually full of biases, reflect contemporary trends and of course a lot of popular culture and they are awfully fine with this. 

And then there is this subculture of creatives, which then really likes to control the generation process from the U-Net to Sampler and who train their own models or write their own extensions. But this group is small, and they use the tools which are already there and which are improving in this regard. From these people, ComfyUI, SwarmUI, even Fooocus get more attention than the mere “wrappers” like Leonardo or Krea. Because they allow granular control, and it's free. 

And Cogniwerk wants to serve both. Ultimately we also aimed to be a hub for every medium, be it text, sound, animation or even 3D. But now there are tools for every medium which specialized in each dimension. Because it took long to implement Stable Diffusion CogniWerk focussed first on image relating tasks. Thus CogniWerk wants to be a Leonardo, a Midjourney or even Adobe competitor, but where do we stand in this regard? How ultimate are we? How creative are we? And what does the Copilot stand for? The name CogniWerk was already generated by an algorithm, and we deemed it, as it was not taken back then, as a good name, because it reflects the intelligent, cognitive part of creating an artwork (in German: “Werk”). And then people come to me and ask me about what is this “coniwerk”. It already starts there. And we tell them that they have full control, and then they fail to find the download button or click on the wrong item. 

So we definitely need to have a way better Copilot, maybe even a Nanny, for our users. And this Nanny has to foster explicit creativity, which means, allow surprises, inspirations, failures, and really tell the user it's okay to try something new. 

And so we do not fight against a Goliath, whose name might be Adobe, Midjourney or OpenAI. We fight against the user's short attention span and the inability to take “creative” risks, to explore, learn and to eventually fail. 

CogniWerk takes from the OpenSource community but is hesitant to give back. We “give” back by making some stuff available easier, with no code and no setup. But in this regard we have to get way better - and faster. If RunDiffusion, Playground or Leonardo train their own model and release it to the community to tinker with, where do we stand? If it takes a whole iteration to implement a new feature, which popped up in the community, while our competitors do it on the fly, where do we stand? If our image quality, generation time and stability is lacking, while even small single use services like cream.ai or styleof.ai can pull it off seemingly easily, where do we stand? 

CogniWerk has to decide what it wants to be. We started as a service run by creatives for creatives. And yet we see creatives do not use us. We can always complain that it is the “users” problem, but complaining won't help us. We have to narrow down and realize that most of the creatives do not want control. They want good results fast. They want to compete with their peers. They want what they see on social media. They do not want the Cogni in the Werk, just the Werk. 

And CogniWerk has to react to this. We can’t just stand here and see how creatives are getting lazier and more amazed the more a tool does for them. We have to put the Cogni back into the Werk, we have to tell creatives, it's okay to struggle, we got you. In the end we have to become our mission: we have to become the ultimate creative copilot.