Feedback on Generative Fill Success Rates and Credit System in Adobe Firefly

[KTDN]

Dear Adobe Firefly Team,

I’ve experienced low success rates with the generative fill feature, using 89 of my 125 credits without achieving desired results. This process can be frustrating and costly. I suggest charging credits only for generations that users decide to keep, which could also aid in improving your model based on successful generations.

Thank you for considering this feedback.

[Jill_C]

I think that would encourage users just to play for hours with the software, never downloading anything, and clogging up the system for users who know what they want and how to get it.

[KTDN]

Dear Jill C.,

I appreciate your concerns regarding the potential for system clogging if Adobe Firefly only charges credits for kept generations. However, I’d like to present some counterpoints based on Adobe’s current credit system and general trends in creative software usage:

1. Credit Allocation and Control: Adobe allocates a fixed number of generative credits monthly based on the user’s subscription type, which does not roll over to the next month. This inherently limits the amount of free usage, preventing system abuse while allowing genuine exploration and learning. Users are aware of their credit limitations and tend to use resources judiciously to avoid running out before the month’s end .
2. Incentive for Quality, not Quantity: Charging for desired outputs could encourage users to think more critically about their inputs, leading to more meaningful usage of the service rather than frivolous experimentation. This aligns with Adobe’s aim for Firefly to enhance creativity, providing tools for serious creation rather than mere play .
3. Impact on System Resources: The computational cost tied to generative AI tasks like those in Firefly is significant. By associating a cost with final products, Adobe can balance server load and manage resources effectively, ensuring a smooth experience for all users .
4. Market Trends and Consumer Behavior: Other platforms with similar usage-based billing have not reported significant issues with system clogging. Users tend to self-regulate based on the value they receive from the service. Additionally, creative professionals value their time and are likely to download only those results that meet their specific needs.
5. Potential Solutions: To address your concerns, Adobe could implement additional safeguards, such as daily or weekly usage caps, to prevent system abuse without hindering genuine exploration and creativity.

In conclusion, a model where users are charged only for results they wish to keep could enhance user satisfaction and the quality of feedback for Adobe, all while maintaining system integrity through existing credit limits and potential additional safeguards.

Thank you for considering this perspective.

[Monika Gause · Accepted solution]

Every request you make, needs to be processed by the servers and that costs a huge amount of money. Currently no AI company earns money with the service. They need to have at least some of their costs compensated. 

[KTDN]

Dear Monika Gause,

Thank you for highlighting the significant operational costs associated with processing each request on Adobe’s servers, an aspect critical to understanding the financial model behind AI services. It’s clear that managing these costs is essential for Adobe’s sustainability in offering generative AI features like those in Firefly.

However, I propose considering a balanced approach that could address both Adobe’s financial sustainability and user engagement:

1. Efficient Usage Encouragement: Implementing a model where users are charged for outputs they choose to keep could encourage more efficient and deliberate use of the service. While it’s true that each request incurs costs, focusing on user satisfaction and value could lead to higher overall engagement and longer-term customer loyalty.
2. Hybrid Credit Models: Adobe could explore hybrid models where users receive a certain number of ‘free keep’ generations per month, beyond which a fee is applied. This method could help cover costs while still allowing users freedom to explore and create without fear of wastage.
3. Cost-Benefit Analysis: While no AI company may currently be profiting significantly from these services, the investment in user trust and satisfaction could result in long-term gains through subscription renewals, upsells, and word-of-mouth recommendations.
4. Incremental Charging Mechanism: Introduce a tiered system where initial explorations are free or cheaper, but as usage increases, so does the cost. This could balance out heavy users consuming more resources while allowing casual users to explore without heavy costs.
5. Community Feedback and Iteration: Adobe could continuously engage with its user base to understand their needs and adjust credit systems accordingly. This would not only help in aligning the product closer to user expectations but also in fostering a community of loyal users.

Understanding the challenges Adobe faces in managing operational costs, it’s crucial to find a middle ground that supports Adobe’s financial health while empowering creators. Your points bring necessary attention to the economic realities of providing such advanced services, and it’s in the spirit of constructive feedback that I suggest these alternatives.

[Adobiwan]

Does Firefly for video have a 'First Frame' preview feature?  That would help!  Show me the first frame first! THEN make the action part of the script happen.   Charge me a dime,, first please,, then the dollar for 5 seconds of action,, if I like the stage! 
I started using a 2nd AI video generator this week, because I burnt up my Firfly $199 this month in just 3 days for my current project (and I'm a hobbyist!! )   I'm doing this part time,, and I spent $199 in three days?   What the?  ----------So,, I'm with KTDN on this.  There needs to be a hybrid path,,