Inspiration
Online D&D campaigns are the source of memorable moments. Why not capture them for posterity using the magic of AI?
What it does
Legend Lore is an automated storytelling pipeline. Discord D&D groups record their session using Craig Bot, then Legend Lore transcribes the recording, selects the best moments, and weaves them into a highlight reel of illustrated story cards + narration. Gemini's interleaved output mode empowers the video to be both narratively coherent and visually stunning.
How we built it
I used Claude Code to generate detailed plans separated into three different phases of development. Then, writing the code was very fast with Opus and Sonnet. Tuning the prompts, thinking about the narrative flow, and doing quality control were the most challenging and time-consuming factors here.
Challenges we ran into
The biggest challenge was prompting Gemini to produce the illustrations and narration according to specifications. Using separate calls for each moment produced more unique illustrations but sacrificed character consistency, so I went with a single prompt. This comes with its own challenges that I am still working out (Gemini likes to reuse portions of illustrations between scenes), but the overall result is better.
Accomplishments that we're proud of
Getting the pipeline to work end-to-end
What's next for Legend Lore
Live video using Veo and expanding the project to work with any D&D group/campaign!
Built With
- assemblyai
- discord
- docker
- ffmpeg
- gemini
- google-cloud
- google-cloud-run
- node.js
- typescript
Log in or sign up for Devpost to join the conversation.