Inspiration

Users may require inspiration to record content for their social media that follows a structured storytelling. By prompting a vague idea for a story, this project will provide a script along with visual storyboards that serve as inspiration for the user to record more narrative video content with the purpose of education, entertainment, or business.

What it does

  • Prompt instructions with guidelines to create a rich and structured script starting from a vague story idea or narrative goal.
  • These guidelines are non-restrictive for creative output, meaning that they take advantage of the creative outputs from Google's AI LLM to create a new coherent script each time.
  • Business owners may record content that with an engaging narrative for their products; social media influencers may record new narratives that entertain their audiences; educators may design a narrative that explains a new concept to a wide audience.

How we built it

  • Structured prompts for specific genres following different story structure designs.
  • Added restrictions to avoid showing fantastical elements that can't be recorded

Challenges we ran into

  • Despite filling out the application form to use Google Imagen's model but I didn't get a response, so I had to opt to use OpenAI's image model. No API Documentation publicly available to incorporate potential Imagen functions for image generation.
  • During AI image generation, character design showed inconsistent elements. AI image models would include new or reinterpret established visual elements (hair-do's, new clothes) . It is difficult to control AI image outputs, but they may serve as inspiration nonetheless.
  • Prompt responses are by default non-deterministic, but the inclusion of a dummy JSON parser with Python would allow their control.
  • Reached token limit multiple times. ResourceExhausted: 429 Resource has been exhausted (e.g. check quota).
  • Reached AI image prompt token limit.

Accomplishments that we're proud of

A GenAI product with No-Code tools from the user.

What we learned

  • Prompt Engineering
  • JSON parsers for stuctured and reusable prompt responses.
  • Basic website Design

What's next for Lights! Camera! Prompt!

  • New Website Design
  • Incorporate Google's Imagen Text-to-Image Diffusion Model.
  • Incorporate multi-modal input. The user may use an image as a story inspiration along with a description.
  • Incorporate Voice narration
  • Incorporate video clips with duration that takes into consideration the voice narration length.
  • Incorporate concepts about cinematic shots for AI Image prompts.

Built With

Share this project:

Updates