-
-
First screen of the app
-
Title Page for the Comic generated using the prompt
-
Page 1
-
Page 2
-
Page 3
-
Page 4
-
Page 5
Inspiration
The human mind is one of the most creative things ever. A major function of the imagination is storytelling-the ability to conjure entire worlds, characters, and emotions from nothing. Yet for most people, that story stays locked inside. You can describe it, but you can't show it. AbsoluteCinema was born from a simple question: what if anyone could transform the story in their head into a cinematic comic book, just by speaking or typing it out?
What it does
AbsoluteCinema takes a story prompt — typed or spoken — and transforms it into a fully illustrated, genre-controlled comic book. Users choose their genre, visual style, and page count. The app then generates a cinematic cover, followed by sequentially illustrated pages with narrated panels, character dialogue, and consistent art style — all powered by Gemini's multimodal generation on Vertex AI.
How we built it
We built AbsoluteCinema using Gemini 2.5 Flash Image on Vertex AI for native interleaved text and image generation. A custom Python pipeline in agent.py orchestrates the story-generation process page by page, maintaining narrative continuity through a rolling story-context window. Google Cloud Speech-to-Text powers the voice input. A Pillow-based page stitcher assembles individual panels into comic pages with text overlays for narration and dialogue. The app is deployed on Streamlit Cloud with GCP service account authentication.
Challenges we ran into
Building with a cutting-edge model like Gemini 2.5 Flash Image came with real challenges. The model occasionally returned empty candidates or fewer images than expected, requiring robust retry logic at both the page and panel levels. Rate limiting (429 errors) under concurrent load forced us to rethink quota management. Getting truly interleaved text and image output — where narration, dialogue, and visuals align panel by panel — required careful prompt engineering and post-processing. Stitching panels into a coherent comic layout with dynamic text overlays using Pillow was an exercise in patience and precision.
Accomplishments that we're proud of
We're proud that AbsoluteCinema actually works — end to end, live, for any story a user can imagine. The cover art quality, the visual consistency across panels, and the cinematic feel of the output genuinely surprised us. We're also proud of the session isolation system that allows multiple users to generate comics simultaneously without conflicts, and the voice-to-comic pipeline that makes the experience truly accessible.
What we learned
We learned that working with frontier multimodal models requires as much defensive engineering as it does prompt craft. Gemini's interleaved output is powerful but unpredictable — building resilience into the pipeline was just as important as building the feature itself. We also learned that storytelling is universal — every genre, from Panchatantra fables to World Cup dramas, comes alive beautifully when given the right visual treatment.
What's next for AbsoluteCinema
The story doesn't end here. Next, we want to add TTS narration so comics can be heard as well as read, motion comic support for animated panel transitions, and a shareable comic export so users can download and share their creations. Long term, AbsoluteCinema could become a platform for educators, storytellers, and creators worldwide — turning imagination into art, one panel at a time.
Log in or sign up for Devpost to join the conversation.