Inspiration
Crumb Canvas™, an animated storybook world I created began as an exploration of playful surrealism and food. With thousands of Midjourney-generated images in my library, I treat my archive as a visual idea engine. One image, a single piece of crying bread, sparked the question: what if bread had feelings? That image became the seed for an entire narrative exploring consciousness.
What it does
When bread across the town of Custard Cove™ suddenly develops emotional awareness, an investigation unfolds to solve a crisis that challenges everything they know about food and consciousness. The mystery leads to an scientific experiment gone...right.
Set in the magical realism of the Crumb Canvas™ universe, this dark absurdist comedy uses documentary-style storytelling to explore the absurdity of existence through the lens of bread with emotions that still fulfills its destiny of being consumed.
How we built it
- Visual Generation: Midjourney for most imagery and animation sequences.
- Supplemental Animation: Kling 2.5 via KREA for select motion sequences.
- Image Editing: Light touch-ups using Nano Banana via KREA
- Audio: ElevenLabs Voice Design. Lip Sync via KREA
- Music: Sound effects and licensed tracks "La Valse de Paris" by Evert Z, "Roma Tarantella" (Instrumental) by Lance Conrad, and "Inspector Bernard" by Louis Adrien from Artlist
- Editing: iMovie
- Workflow: Solo production combining AI-generated visual assets with traditional filmmaking techniques
Challenges we ran into
The biggest creative challenge was figuring out how to handle character consistency. I'd been experimenting with multiple visual styles in Midjourney, creating slightly different aesthetics for different characters and scenarios. Training a LoRA to lock everything down felt like the proper solution, but it would have forced visual uniformity across a story that didn't need it.
The breakthrough was realizing the style variation could be a feature, not a bug. By leaning into documentary-style storytelling, each character's perspective could have its own visual treatment. Pete the scientist's worldview looks different from Tilda's playful precociousness because they see the world differently. The multi-media aesthetic became the storytelling mechanism rather than a technical compromise.
Accomplishments that we're proud of
I made deliberate aesthetic choices that run counter to current AI filmmaking trends. While many productions chase photorealistic polish and technical perfection, Shelf Life embraces visual absurdity and mixed-media storybook illustration.
The multi-media approach came from my style experiments in Midjourney. I'd developed several distinct visual languages, and rather than forcing consistency, I used the variation as a storytelling tool. Each character's section feels visually distinct because their perspective is distinct. That choice required the narrative to be strong enough to carry viewers through intentional stylistic shifts, but it made the film more interesting than locked-down visual consistency would have.
What we learned
The beauty of adapting your story to the reality of the tools! There are three stories: The one in your mind, the one you generate and the one you edit. You are in a good place if they overlap somewhat. AI tools are collaborative partners, not servants. My best results come from dialogue with the technology rather than control over it.
What's next for Shelf Life
Shelf Life is a stand alone feature of the Crumb Canvas™ universe I created but the characters are a fixed part of the world with their own storylines in development.
Built With
- artlist
- elevanlabs
- kling
- krea
- midjourney
- nanobanana
Log in or sign up for Devpost to join the conversation.