In our recent Pixelz AI Unveiled webinar, our two in-house AI experts, Janus Klok Matthesen and Sébastien Eskenazi, answered the most frequently asked questions from our community about the AI models at Pixelz.
With nearly a decade of computer vision experience, Sébastien has led Pixelz's R&D team, assembling an AI automation powerhouse, while Janus has led Pixelz's research and development teams, which have developed some of the most advanced AI and software tools for photo retouching.
Sébastien and Janus went into more depth to the questions about AI in our recent webinar, but we have concluded a few important questions and answers from it down below.
If you want to learn more about the topic, we encourage you to check out the full webinar here.
How Pixelz AI Works and Can You Use Different AIs to Generate Product Images?
At Pixelz, we have a production system that we call S.A.W., where we split the process of editing an image into different steps. Some of these steps are done by human retouchers and others by our AI models.
Superior quality, mass production, personalization, and fast delivery times work against each other on conventional methods, but with Pixelz’s AI tools, these four attributes begin to work together and complement each other.
While our customers may not be aware of what the AI tools are doing on the back end, the benefits are visible in:
- High-quality success rate
- On-time and fast delivery
- Real-time and transparent image status
On top of Pixelz’s built-in AI tools, many organizations are considering how generative AI image generators like Midjourney or DALL-E by OpenAI could help create product images.
As of now, there aren’t any great AI image generator solutions that wouldn’t change how the product looks. Currently, though, there is a lot of exploration around how generative AI solutions can change the background in product images and assist the creative process. At Pixelz, we don't use generative AI tools to create product images themselves but are looking forward to how else generative AI can be used in studio workflows.
How Does the Cost of AI Compare to Traditional Post-production?
The cost of AI depends because the financials are different compared to traditional post-production. This is how they differentiate from each other:
- Traditional post-production: It has a low entry and setup cost, as you can hire people, and they get to work. In the long term, you can’t make as many savings with it as with AI.
- AI models: If you want to build an AI model from the ground up, you're going to need to invest tens of thousands, if not even hundreds of thousands of dollars, to just build the model. In the long run, AI is much cheaper to run, but the upfront costs are much higher.
However, since there are existing AI tools on the market, photo studios don’t have to train an AI model from the ground up. For this reason, compared to traditional methods, implementing AI can be a significantly cheaper long-term option.
For instance, with jewelry products, it can take you an entire day just to shoot one watch, but with combining CGI and AI, your team could get the product photos produced much more efficiently. If you want to look for a direct example, one popular brand that does this with their product photos is Van Cleef & Arpels.
Creating these AIs and automation is what has enabled us at Pixelz to create the product we call Flow, where we continuously get images for retouching from our customers. We can then deliver them back with as little as one hour of turnaround time.
For many of our customers, a fast delivery time is a must, and using AI models enables this fast and cost-efficient service. If we use traditional retouching techniques without knowing the volume of images and suddenly get thousands of images, we simply can't deliver them back within an hour.
With AI, it's a matter of how many computers we turn on and start running. AI enables scaling that wouldn't be possible otherwise.
