This week, I had a terrible experience with a pretty promising product. There were a ton of product lessons in my experience, but in this article, I want to focus on major new shift in pricing: Consumers paying for compute resources.

It used to be that only engineers working with AWS or GCP had to think about how much time and cost was associated with compute time. But in the age of AI, compute costs can make or break a start up.

I am seeing two major models:

• Monthly Subscriptions: ChatGPT, Midjourney, etc.

• Small Subscription with Point System

There are other modifications with more transparent Pay-As-You-Go (PAYG) systems, but I can’t think of any mainstream pricing model that is just PAYG.

For this article, I will ignore the other problems this product had. I just want to focus on the big debate around pricing for GenAI as we know it today.

Companies like OpenAI (ChatGPT) and Midjourney are figuring out their costs and letting you buy a clear package that includes an easy-to-understand list of items like # of requests processed and the speed or size of files processed. If you run out of room, you can upgrade.

What I discovered this week was a different model that puts a horrible level of cognitive load on the end user. I found myself flashing back to when we moved from the software ownership model to the subscription model (Think about Adobe Photoshop and all the people — myself included for a long time — who were frustrated with having paid for software and then suddenly being told, now you have to start paying monthly fees for a new improved product we didn’t want/need/understand).

The current shift is from subscription models to subscriptions with a pay-as-you-go component. In this case, I am talking about Fliki.ai, and there pricing system that is sooo complicated you have to navigate a complex pricing model that reminds me of a Starbucks coffee menu for those of us who don’t drink (fancy) coffee.

The Pricing Menu

The subscription model requires you to understand the “credit estimate” and how to “optimize” it. Stop. Take that in for a minute. When I subscribe, I need to know roughly what features I want and need and how often I am going to use them. Someone should seriously think about an AI driven tool or a good old fashioned spreadsheet to help me work out what my credit budget will be based on my needs.

I get that they are breaking new ground but there is a greater than 50% chance they will fail on their pricing model alone. Their layout is really amazing, but a larger company could price more sensibly and just mimic their key features and be done with them.

This reminds me of sorting through health plans in November every year and trying to decide: how much healthcare do I think I need. I hate this pricing model as much as I hate trying to figure out health insurance plans based on usage I am not sure about.

Credit Estimate

The standard plan comes with 180 mins of credits per month.

Ideally, you could create 90 mins of videos per month (containing audio). A lot depends on how much credits would be spent on creating/editing the audio.

Most users easily get 45-60 mins of videos created using 180 mins of credits.

Fliki offers tremendous value as outsourcing video creation for a 2-3 min video with voiceover would easily cost around $30-$50. (20x more expensive).

Tips to optimize credits

Keep sections small: Having shorter sections (1-2 sentences) will make sure you can edit the text without re-consuming extensive credits at every edit.

Use Pronounce: If you want to listen to how a particular word sounds without playing the entire section, highlight the text and select “Pronounce” to listen to just the word. And you can make corrections in the Pronunciation map.

Review before Preview: Please review all your script and edit it before you hit play and preview the audio/video. This will save an ample amount of credits.

Summary

It will be interesting to watch as pricing policies designed for engineers are pushed out to the general public. Creators are good at what they do, but unless their creations are math and/or accounting the aforementioned pricing model is likely doomed to fail. Fliki.ai might be on the cutting edge, or our need for AI commute so crucial that we are willing to endure Healthcare-like pricing models. But odds are, startups need to figure out better ways to build subscription plans.

From a product usability stand point, the pricing model doesn’t allow the average user to achieve an aha moment before being asked to fork over $30. This will keep a lot of good customers out. For me, I caused me to find a better solution that is more-inclusive with more sensible pricing. If it solves my problem, I will pay more but at least I can reach the aha moment before paying more than is comfortable.