Skip to content
K

Kahani

Let your characters and stories take flight.

Created on 9th November 2025

K

Kahani

Let your characters and stories take flight.

The problem Kahani solves

The Problem with Children's Screen Time
Our platform directly addresses the key challenges parents, educators, and children face with modern digital content.

The "iPad Kid" Problem: Passive Consumption & Stifled Creativity
The Problem: Children's screen time is increasingly passive, consuming low-effort, algorithm-driven content. This trend, often called "iPad Kid" syndrome, leads to mental stagnation ("brainrot") by replacing imaginative, unstructured play with non-interactive consumption. As a result, creativity and critical thinking skills are not being developed.

Our Solution: We directly reverse this. Our platform is an active tool that demands creativity first. By requiring a child to draw, we spark their imagination. By instantly rewarding that drawing with a custom cartoon, we turn them from passive consumers into proud creators.

The Parent-Child "Bonding Gap"
The Problem: Parents and children often struggle to find a shared activity that genuinely engages both. Kids are drawn to digital worlds, while parents seek quality, developmental time. This creates a "bonding gap" where shared experiences are missed.

Our Solution: We bridge this gap by creating a perfectly collaborative activity. The child provides the art (their strength), and the parent provides the lesson (their guidance). They work together to create a unique story, fostering a shared sense of accomplishment and a powerful bonding moment.

"One-Size-Fits-All" Educational Content
The Problem: Generic cartoons cannot address a specific child's real-time developmental needs. A parent struggling to teach their child about sharing, potty training, or eating vegetables has no tailored media to help them.

Our Solution: We provide on-demand, personalized lessons. Parents control the narrative, allowing them to instantly generate a high-quality cartoon that teaches the exact academic or moral lesson their child needs that day.

The Special Needs "Content Gap"
The Problem: Children with special needs, such as autism or ADHD, require highly specific content. Autistic children benefit greatly from "social stories" to model behavior, while children with ADHD require dynamic, engaging visuals to maintain focus. This content is expensive and difficult to find.

Our Solution: Our platform is an ideal tool for these exact needs.

For Autism: Parents can instantly generate unlimited, personalized social stories starring their child's own familiar character, making the lessons less intimidating and more relatable.

For ADHD: The bright, dynamic, and fast-paced nature of custom animation is perfectly suited to capturing and holding the attention of children with shorter focus spans.

The "Empty" Character Interaction
The Problem: Even in "interactive" apps, characters are often just superficial animations. There is no real, dynamic relationship.

Our Solution: By integrating Gemini Flash (for chat) and Gemini Live (for calls), our characters have a persistent, interactive personality. The child can create their hero in a cartoon, and then talk to that hero in real-time, building a genuine, engaging, and educational friendship.

The Character "Drift" Problem (Technical)
The Problem: A major technical hurdle in AI video generation is character inconsistency. Most models "drift," causing a character to look different from one scene to the next, which breaks the magic and immersion.

Our Solution: Our I2V (Image-to-Video) pipeline, powered by Nano Banana and Veo, solves this completely. By generating a static, consistent keyframe for every scene, our hero looks 100% identical from the beginning of the cartoon to the end.

Challenges we ran into

Challenges We Ran Into
We encountered two significant hurdles during this project. The first was a technical bug regarding video consistency, and the second was a product-level challenge in transferring context between the different AI models.

1. The "Character Drift" & "Scenery Jump" Bug
Our biggest technical hurdle was ensuring absolute character and scene consistency across the four 8-second video clips.

The Bug: Our initial approach was a simpler Text-to-Video (T2V) pipeline. We used Gemini 2.5 Pro to generate 4 sequential T2V prompts for Veo, (e.g., Scene 1: Hero in a messy room..., Scene 2: Hero sees a monster in the messy room...). The result was a failure. Veo, treating each prompt as a new task, would re-generate the scene from scratch every 8 seconds. This caused a jarring "jump" where the hero's clothes would change, their facial features would "drift," and the room's layout would reset. It completely broke the illusion of a single, continuous cartoon.

How We Got Over It: We re-architected our entire pipeline to an Image-to-Video (I2V) workflow, making character consistency our top priority.

Stop Telling, Start Showing: We stopped asking Veo to imagine the scene and instead tasked it with just animating a pre-existing image.

Use Nano Banana as the "Keyframe Artist": We now use Gemini 2.5 Pro to generate 4 image prompts for Nano Banana. We explicitly instruct Nano Banana to place the same character avatar into 4 sequential keyframes, critically adding (maintain scenery from previous image) to the prompt.

Use Veo as the "Animator": This gives us 4 static, 100% consistent keyframes. We then feed these 4 keyframes into Veo with simple animation prompts (e.g., Animate this: The hero picks up the toy), not generation prompts.

The Result: By animating the existing pixels rather than re-imagining them, we completely solved both the "character drift" and "scenery jump" bugs, resulting in one seamless 32-second video.

2. The "Amnesiac Chatbot" Hurdle
Our second challenge was connecting the generated cartoon to the interactive chatbot. After a child watches their 32-second cartoon, they can chat with the hero (powered by Gemini Flash) or "call" them (powered by Gemini Live).

The Hurdle: By default, the Flash/Live models had no "memory" of the cartoon that was just created. If the child said, "Wow, we defeated the Grime Goblin!" the chatbot would respond with, "I'm sorry, I don't know what you're talking about." This was a major product failure that broke the entire experience.

How We Got Over It: We used Gemini 2.5 Pro as the central "brain" or "state holder" for the entire experience.

Generate a "Persona Summary": After Gemini 2.5 Pro generates the 4-scene script, it also generates a dynamic "persona and memory" summary.

Example Summary: [SYSTEM PROMPT FOR FLASH/LIVE]: You are 'Broccoli-Hero', a brave hero who loves vegetables. You just had an adventure where you defeated the 'Sugar-Slug' by eating your carrots. You are feeling brave and proud. The child you are speaking to is your best friend who helped you.

Dynamic Prompt Injection: This summary is automatically injected as the initial system prompt for the Gemini Flash and Gemini Live instances.

The Result: When the child starts the chat, the AI already knows its name, its personality, and the adventure it just shared. This creates a seamless transition from passive viewing to active, in-character interaction, making the experience truly magical.

Tracks Applied (1)

Best Use of Gemini API

Our project represents a "full-stack" implementation of Google's generative AI ecosystem. We are demonstrating a sophist...Read More
Major League Hacking

Major League Hacking

Discussion

Builders also viewed

See more projects on Devfolio