Interview with the founder of Zust AI - Saquib

Interview with the founder of Zust AI - Saquib

Can you share the inspiration behind venturing into the generative AI field, and what motivated you to build Zust AI?

I have been tracking generative AI for a long time.
GANs came out around 2014–2015. Back then, image generation was very limited, mostly around generating faces and images in different styles. As we already know, things started changing for generative AI by the end of 2022. At that time, me and my co-founder were working on a different idea related to educational content creation while traveling around India.

That idea did not pan out as we had expected, so we decided to switch gears and start exploring ideas in generative AI. It did not take long to narrow down on image generation space. So we quickly built a prototype over the next few days and asked Chat GPT to find a name that represented speed and agility, and that is how zust was born. The idea behind zust is to enable everyone to create personalized photos and videos quickly and effortlessly.

What challenges did you face while developing Zust AI, and how did you overcome them?

So, we were specializing in product photography; therefore, the utmost priority was to reduce deformations in the product's image to zero.
This was a big challenge in the initial days. We used outpainting back then, which was not as creative as text-to-image models. Plus, it would always deform the product. We had a break-through when ControlNet came out. Currently, we use a combination of models to guide image generation.

What advice would you give to aspiring entrepreneurs looking to enter the generative AI space and build their own applications?

Implement fast. Try not to lock-in into any particular model. avoid training your own model initially, engineer the platform first using the open-source models and APIs. We have started using fine-tuned models now.

In the rapidly evolving field of generative AI, how do you stay updated with the latest advancements and ensure that your applications remain cutting-edge?

Be active on Twitter and arxiv.org, seriously! At zust, we try to test out new paper/method as soon as possible and analyse its application in our context. Also, we try to use API and open-source model everywhere.

Can you share a memorable success story or milestone from your journey as an entrepreneur in the generative AI field?

Stable diffusion is great and creative, but Midjourney is far ahead in terms of image quality and composition. So, we had this idea of personalising midjourney images in just few clicks.
We build the stack over a weekend and released it on producthunt, to become #5 product of the day. The response from the community was amazing, and it got monetised on the first day of the launch. It was exciting to see my creation getting monetised so quickly. This was in October.

Finally, a few words about you so people get to know you better, and how to reach you for questions or feedback about Zust AI.

At zust.ai, we help brands to market their products faster by accelerating product photography and social media content creation using generative AI. We have clients from furniture, jewellery and FMCG industry. Reach out to me at saquib@zust.ai for any query related to collaboration and product demo.