About AI Reasoning, Image Generation features (Bonkers) & Future Features
under review
Gabriel Hoffmann
Dear Merlin Team,
I have a few questions regarding the AI models in Merlin, particularly DeepSeek, ChatGTP, and others, which include an integrated reasoning function. Unfortunately, I am unable to enable or disable the "reasoning" function in Merlin. Could you please explain how I can access it?
I also have some questions about image generation:
Will the image creation tool "Bonkers", which is still in development, be included in the Merlin subscription once it is released? And will it also be available for AppSumo Tier 3 buyers?
Additionally, I would like to know whether it will remain possible to adjust essential parameters such as the sampler selection, negative prompts (Stable Diffusion), the number of image generation steps, and creativity (prompting guidance). These functions are crucial for me and many other users to ensure flexible and precise image generation.
Furthermore, I am curious whether AI video generation will be integrated into Bonkers/Merlin in the future. Are there already plans or a roadmap for this feature?
Thank you very much for your support! I look forward to your response.
Best regards
Hoffmann
endu
under review
Kunal Verma
Hey Gabriel Hoffmann,
- Reasoning is available in a few models (examples include Deepseek R1, Claude 3.7 Sonnet, O1, and O3 mini), so when you select these models, reasoning is enabled by default.
> Will Bonkers be included in the Merlin subscription in the future?
Currently, yes, Bonkers is included in the Merlin subscription. However, we might introduce separate subscriptions for Bonkers down the line.
> Need more parameters
- Are there specific use cases you're exploring on Bonkers?
- Are there scenarios where the generations/edits don't meet your expectations? If so, please share examples where greater control over these parameters would make a significant difference.
> Video generation?
Yes, we are planning to add video generation and editing capabilities to Bonkers.
cc: Siddhartha
Gabriel Hoffmann
Kunal Verma
Dear Kunal Verma,
Thank you for your response. I have a few additional questions and suggestions regarding Merlin, Bonkers, and future software development:
- Reasoning Function: Activation and Visibility
You mentioned that Reasoning is enabled by default. However, responses (e.g., via DeepSeek/ChatGPT) are generated so quickly that no discernible "thinking process" occurs—even with complex queries. When I input the same task on the models’ official platform, the process takes significantly longer.
Question: Is there a way to visualize the Reasoning process or verify if it is truly active? It seems unlikely that a full Reasoning process occurs at the current speed.
- Bonkers Access and Merlin Subscription
It is currently unclear how Bonkers will be integrated into the Merlin subscription moving forward:
Will Bonkers features be entirely removed for Merlin subscribers, or will generation costs (credits) likely increase?
Are there concrete plans to retain AI image generation in the subscription? Losing this would be critical, as it is explicitly part of Merlin’s offering.
You also mentioned future video generation. Will this be integrated directly into Merlin, and how will it link to Bonkers/subscription models?
- Missing Features in Image Generation
To make Merlin appealing to advanced users, the following features are missing (essential for Stable Diffusion models):
Seed Control: Random seeds or repeating the same seed.
Prompt Guidance: Adjust how strictly the AI adheres to prompts vs. creative freedom.
Step Adjustment: Quality control via the number of generation steps.
ControlNet Functions: Precise control over image content.
Sampler Selection: Options like Euler-A or DPM++2M-Karras.
Prompt Book: Storage for frequently used prompts.
Image Deletion: Option to delete unwanted images (currently missing this basic feature).
Image Library: Organize images into custom categories/folders (e.g., with tags/filters).
Suggestion: Advanced settings could be "hidden" for casual users (e.g., under "Expert Mode"). A library would greatly optimize workflows.
I look forward to your insights and appreciate your efforts in advance!
Best regards
Gabriel Hoffmann
S
Siddhartha
Hey Gabriel Hoffmann, can answer for the subscription part. There will always be some capability to generate images in merlin. (In chat you can generate images if you prompt it to). For specialized tools, we will probably take it out of the standard subscription in the future.