Team: conceptual_slipknot
Members: Soumya, Sashank, Bill and Mistral AI
Submission for the London Mistral Hackathon
Co-Author is designed to be a creative criticed multi-modal co-pilot for authors, enhancing the literary crafting process by generating poetry based on images and themes. Think of it as a control-net for transforming visual art into textual creativity!
-
Image Input:
- Feed an image into the system to identify significant objects using Pixtral.
-
Theme Extraction:
- Derive a central theme from the analyzed image.
-
Contextual Significance:
- Understand the significance of the extracted objects in relation to the theme, providing context for creativity.
-
Poem Generation:
- Generate an original poem based on the identified significance and context.
-
Image Handling:
- Introduce a new image at any point.
- The system allows for image replacement to keep the creative process dynamic.
-
Iterative Process:
- Repeat Steps 2 & 3 as necessary, refining the theme and context.
-
Critique Mechanism:
- A LLM-based critic assesses the relevance of any new additions.
- If the new image or poem doesn’t align with the context or quality, it can be regenerated.
- Note: No critic is applied to the first poem generation.
-
Dynamic Adaptation:
- Both the poem and the image can be modified at any point, ensuring relevancy and coherence throughout the creative journey.
run run
:P
- To utilize the Mistral API:
pip install mistralai