Understanding our New Pricing Model for Additional Resource Usage
We’re updating the pricing model for image generation with additional resources (LoRAs, DoRAs, LyCORIS, Embeddings, etc.) to better reflect the costs incurred and resources required to generate your images.
Here's a breakdown of the changes and what they mean for you;
Why the Change?
Each time you add an additional resource to a generation, it increases the demand on our system’s “Workers” (the GPUs that generate your images). While popular resources are kept loaded on our Workers for quick use, unique combinations require extra processing. When you request a combination of resources that haven't been used together recently, our system has to load them onto a new Worker, which costs both time and money.
With hundreds of thousands of LoRAs, embeddings, and other resources available on Civitai, this leads to millions of unique combinations. Each unique set requires extra steps, and the cumulative effect of this is significant. To keep Civitai running smoothly and fairly for all users, we need to adjust pricing to account for these costs.
Civitai’s financials are no secret - we’ve been completely transparent about our challenges, and our 2024 Transparency Report lays it all out for anyone to see. The reality is, we're not profitable and continue to operate at a loss while working hard to keep the platform sustainable for everyone. Every decision we make, including this small price adjustment, is about ensuring Civitai can continue providing the tools and resources that our community relies on. If you haven’t seen it yet, you can read the full report here to understand exactly where we stand.
How Does the New Pricing Work?
We’ve introduced a formula that adds a small cost for each additional resource you use. The more resources you include, especially unique combinations, the higher the cost, as these require extra processing resources. Note that Buzz, behind the scenes, is calculated with a decimal but rounded to the nearest whole number for display.
The formula is:
Added Cost (Buzz) per image = 0.5636713 - 0.01146222 x + 0.07849689 x²
(x = number of additional resources)
Here are a few examples to show how it works:
Simplified Breakdown:
1 to 3 resources: Added cost is 1 Buzz
4 to 5 resources: Added cost is 2 Buzz
6 resources: Added cost is 3 Buzz
7 resources: Added cost is 4 Buzz
8 resources: Added cost is 5 Buzz
9 resources: Added cost is 7 Buzz
10 resources: Added cost is 8 Buzz
11 resources: Added cost is 10 Buzz
12 resources: Added cost is 12 Buzz
What This Means for You
Minimal Impact for Low Additional Resource Usage: If you use just 1 - 3 resources, you'll notice a marginal cost increase; 1 Buzz.
Based on our analysis of generator usage, more than half of users will only see a 1 Buzz Increase.One thing to note is that while we've always provided Civitans the means to stack multiple additional resources, including LoRAs, the most popular, collected, and reacted-to images rarely use more than two LoRAs! In many cases, stacking LoRAs provides diminishing returns for image quality!
Higher Costs for Multiple Additional Resources: As you add more resources, especially when using unique combinations, the cost increases more significantly to reflect the additional processing required for these generations.
All Additional Resources?: This adjustment only applies to additional resources over 10 megabytes in size, excluding VAEs. Smaller files, including 99% of embeddings, will not incur any added Buzz cost.
In the generator you'll see a small weight icon and tooltip next to resources on which there is an additional Buzz cost applied.
Featured Models - Zero Added Cost! We’ve excluded any currently featured model from the added cost calculation! If you see something cool in the Featured Models homeblock on the homepage that you want to try out, you won’t incur any additional resource costs while that model is featured! Want to get your model (or image) boosted to the homepage? Click "Get Showcased" from the Buzz Dashboard!
The Bottom Line
As with any change to pricing, this adjustment isn't something we're implementing lightly.
With millions of possible LoRA combinations and increasing costs due to the thousands of new resources added to the platform each week, this change is necessary to ensure we can continue offering the best possible experience for everyone, while maintaining a sustainable platform.
We appreciate your understanding! See you in the Generator!