From Midjourney to Firefly via Topaz AI
Three tools have been shortlisted to handle the process : first of all using Midjourney AI on Discord Version as it appeared to be the most consistent and reliablecreative iteration tool of it's time, but with the following limitations (obsolete by now) of delivering only low resolution assets (less than 2 megapixels) and not (yet) allowing local adjustments on each iteration.
To properly nurture the learning curve of the AI a pro license was required, allowing to reset and sharpen at will the learning curve of the AI model, based on actual primary sets of photos physically made in our studio in Brussels. Two sets of pictures were therefore combined : a first one destined to teach the model what type of assets we expect while describing objects such as "pineapple slice", "banana chunk" or "little carrot with leaves". Then we produced a second set of pictures picturing the juice cup with the actual receipes combinations around it.
The Half-Cut Cup
Finally a third bunch of pictures were made specifically to explain to the model how the container (paper cup) would be cutted to reveal the inner mix and showcase the client logo in a clean and consistent framing.
Iterating & rewarding the model when the liquid action is consistent with the guidelines
Then the iterations began (600 on average) to create a consistent art direction into the model for each receipe. The validated outputs were then upscaled to 12 megapixels using Topaz Gigapixel AI allowing a high quality postprocessing using Adobe Neural Engine and Adobe Photoshop Ai Assistants for local corrections (now called Firefly).
At Delivery, all assets are downscaled to 2Mpx again and compressed to optimize loading time and smooth experience using the touchscreen devices.
The Result
From brief to Production Delivery the actual full process took places in matter of weeks instead of months - and avoided a significant amount of studio cleaning and potato-tube launcher reloading, even if it sounds funny it can become a real assistant killer when reproduced hundreds, if not thousands of times. Along the full process this included more "traditional" operations for packshot photography workflows such as processing branded shiny metal camping bottles, extra cup sizes, side dishes and other color variations based on a simple but important pre-production step consisting of measuring the color and density of each mix and creating a trustworthy reference palette.
In fine, even if the full process seemed to took the same amount of energy as if it had been produced in a more conventional way, it's effectiveness and the ability to generate new assets almost immediately worth the price of the effort. For the future, this kind of workflows may be even more improved by using cycle of iterations assisted at other steps such as the prompting (involving GPT-4) and the opportunity to animate the results (such as custom workflows available in Runway AI).
