This content originally appeared on DEV Community and was authored by SS
__This is a submission for the Google AI Studio Multimodal Challenge
gemini-2.5-flash gemini-2.5-flash-image-preview imagen-4.0 imagen-3.0
Inspiration
I always wanted to draw comics that can capture my chaotic imaginations – but the drawing, erasing, starting again is such a drag!
Also, AI didn’t help much – create – frustrate – regenerate – repeat and yet couldn’t get my vibe… even more drag!
Well that was until Gemini nano banana
gemini-2.5-flash-image-preview!
I am so blown away by its editing capabilities specially working with multi-image, multi-modal inputs that I couldn’t allow my lazy self to procrastinate anymore !
So, here’s
What I built
SONICS.ai is a comprehensive, a Google AI-powered creative suite 
(demo) that transforms a user’s simple idea into a fully-realized, multi-sensory, character-consistent comic book experience with podcast playbacks,
while allowing them to add their flavours/ vibes at every aspect, from storyline to characters to scenes to dialogues to text styles all in natural language.
The best part?
You dont need to be good at drawing! AI solves it for you.
And you can still bring your creativity, your imagination, your flavour/ vibe, your stories to life using SONICS.ai
without losing your patience with back-and-forth regeneration to get that perfect shot!
You can make full-comics, your style in MINUTES instead of months without lifting a pencil !!
For lazy fellas, you can hear your generated comics !
Demo
My project in action
Multimodal workflow architecture
My project in action
0:00 Intro
0:10
0:20
0:53
1:24
1:34Audio preview
1:47Play the Comic that speaks your Style
Multimodal workflow architecture
gemini-2.5-flashgemini-2.5-flash-image-previewimagen-4.0imagen-3.0
How I Used Google AI Studio
This app was entirely built on Google AI studio vibe-coded from scratch
as you could have guessed by now for my lazy vibes !
I started with a simple idea prompt and kept on adding features by guiding the AI through pain-points I have faced when vibe-creating comics with my flavour.
The Multimodal capabilities I implemented …
Multimodal Capabilities
Input
Output
Models
Feature
Text
Image
gemini-2.5-flash-image-preview
imagen
For quality Character, Scene Background generation
Text editor based updates
Image + Text
Text
gemini-2.5-flash
Automatic character description updates for natural language based character edits
Image (mask) + Image + Text
Image
gemini-2.5-flash-image-preview
For precise edits in characters/ scenes, dialogue corrections, text stylings, positional edits, detail improvement
Multiple Images + Text
A composite image with rendered text
gemini-2.5-flash-image-preview
For comics scene panel generations ensuring character consistencies across scenes, dailogue accuracy, scene quality
Multimodal Features
The specific Multimodal functionalities I built and why it enhances the user experience (UX)…
click this for modality implementation details & respective models before proceeding
1. Composite scene panels
Models :imagengemini-2.5-flash-image-previewgemini-2.5-flash
The comic panels are created through an intelligent composition logic combining the multimodal capabilities of the models to create final panel images from the inputs – scene background, character images, scripts that were themsleves generated by using either of these.
This ensures character consistency, dialogue accuracy as well as scene quality across comic scenes.
2. Flavour edits
Model :gemini-2.5-flash-image-preview
It is used for precise surgical editing of scenes, characters, dialogues, styles leveraging masking.
Users can simply explain their edits in natural language for feature changes (with / without masking).
This helps users avoid regenerating back-and-forth images from scratch which was really frustrating when we need to make a small style/ error correction. And users can add their vibes/ flavours/ styles to the scene.
Acknowledgement
Google AI studio is phenomenal at vibe-coding. I was able to generate and finish a well-working prototype in less that 6 hrs.
But as you could have guessed – Parkinson’s law took most time !
gemini-2.5-flash-image-preview(Gemini nano-banana) is the star of my whole idea. Due to nano banana, I was able to successfully create a consistent character comic experience,
imagenhelped me create beautiful backgrounds for the comic scenes which were then fully realised using composite logic.
gemini-2.5-flashhas been used for prompt engineering for inputs to other models, and also for optimising the deliverables.
Thank you!
It was a fun and great experience!
What Definitely Not a drag!
This content originally appeared on DEV Community and was authored by SS