C6 Fun Writing - English Writing to Creative Visuals

Fun Writing is a creative platform where students practice English writing on diverse topics, get AI feedback, and turn their work into images or videos.

Comment

C6 Fun Writing Project Story

💡 Inspiration

Driving student to practice English Writing is a pain point. English writing requires students to provide different details, which can often feel like a tedious assignment rather than a creative outlet. This lack of engagement directly impacts the quality and frequency of practice. Victor's solution came from the good idea to transform these writing practices into a fun, gamified experience using Generative AI to immediately reward effort. By using the student's own words to generate unique visual media (images and videos), Victor created a powerful, personalized incentiveturning a written assignment into a creative, shareable production.

🎯 What it does

C6 Fun Writing is a gamified web application that converts a student's creative writing into unique, AI-generated images and videos. It manages the entire creative writing process from prompt to portfolio.

1.  AI-Generated Prompts: Provides unique, age-appropriate writing prompts across 20 diverse themes tailored to students aged 3-16+.

2.  Automated Feedback & Scoring: Submissions are instantly analyzed by the FeedbackAgent across four key dimensions (Grammar, Spelling, Relevance, Creativity), with scores from 0-25 points each for a total of 0-100.

3.  Gamified Rewards: Students earn 3,000 starting AI credits. Image generation costs 100 credits per image with 4 styles (Standard, Comic, Princess, Manga). Video generation costs 500 credits per video with 2 styles (Animation, Cinematic), producing 8-second 720p videos.

4.  Digital Portfolio: All stories, feedback, scores, images, and videos are saved to a personal portfolio for easy review and sharing.

🏗️ How Victor Built It

Victor built C6 Fun Writing entirely on the Google Cloud Platform (GCP) to ensure scalability, cost-efficiency, and rapid deployment.

  • Front to Back: Victor utilized a modern React/TypeScript frontend hosted on GCS/Cloud CDN and a Node.js/Express.js backend running on Cloud Run for the main API endpoints.

  • AI Agents & Real-Time Processing: The core intelligence is a suite of 5 AI Agents (ContentSafetyAgent, ImageSafetyAgent, PromptAgent, FeedbackAgent, VisualMediaAgent) deployed as a Cloud Run HTTP Service using Python + Google Agent Development Kit (ADK) + FastAPI. All agents are accessible via HTTP REST endpoints for real-time, on-demand execution with direct integration into the backend API.

  • Generative AI Models: Victor orchestrates multiple models: Gemini 2.5 Flash for text generation (prompts, feedback, visual descriptions, and image safety validation), and Veo 3.1 Fast Preview for video generation.

  • Data & Storage: Victor uses Cloud SQL for PostgreSQL for secure student data (submissions, scores, credits) and a public GCS Bucket (fun-writing-media-prod) for storing the final, unique visual outputs.

🔧 Key Challenges and Solutions in Multi-Modal AI Deployment

1. Orchestrating Multi-Modal AI Pipelines
The media generation workflow follows a multi-step transformation:
Text → Visual Prompt → Image → Video.
Victors Cloud Run HTTP Service had to ensure that the VisualMediaAgent could reliably:

  • Invoke Gemini 2.5 Flash to generate a creative visual prompt from student input
  • Use Gemini 2.5 Flash and Veo 3.1 for media generation to produce image and video assets
  • Upload results to Google Cloud Storage (GCS)
  • Handle API failures and retries with robust error management

This orchestration demanded resilient, fault-tolerant logic to maintain reliability across asynchronous, multi-agent calls.

2. Real-Time Credit Management
Because image and video generation consume user credits, Victor implemented:

  • Atomic credit deduction before invoking costly generation APIs
  • A refund mechanism for failed generations to ensure fairness
  • Real-time validation to prevent race conditions and double charges

This system preserved trust and ensured a smooth user experience.

3. Safety and Moderation Enforcement
To maintain a safe and positive environment, the SafetyAgent served as a proactive gatekeeper by:

  • Filtering user submissions and AI-generated content (prompts, feedback, visual descriptions)
  • Enforcing strict moderation against prohibited topics
  • Ensuring all outputs aligned with platform safety standards

This layer of protection was essential for fostering a creative, inclusive space for students.

💡 Accomplishments that Victor is Proud of

Victor created a web app which can help student to do more writing and having fun using their writing to generate image and video. Beyond the core feature, Victor is proud of:

  • Immediate, Actionable Feedback: Deploying a working, real-time 4-dimension scoring system (Grammar, Spelling, Relevance, Creativity) that gives students instant, meaningful guidance on their work with detailed feedback and next steps.

  • Multi-Agent Real-Time Processing: Architecting 5 autonomous AI agents (ContentSafetyAgent, ImageSafetyAgent, PromptAgent, FeedbackAgent, VisualMediaAgent) as a unified Cloud Run HTTP Service with Python + Google ADK + FastAPI, enabling real-time processing with direct HTTP REST endpoints.

  • Full End-to-End Visual Workflow: Successfully integrating Gemini 2.5 Flash and Veo 3.1 Fast Preview to deliver multi-modal media generation (4 image styles + 2 video styles) to create unique, compelling, and full-circle creative rewards for students.

📚 What Victor Learned

Victor learned how to use Google Agent Development Kit (ADK) to build an efficient, multi-agent AI system. Using ADK with Python + FastAPI, Victor successfully orchestrated 5 specialized agents that work together in real-time via HTTP REST endpoints. This approach demonstrated a robust pattern for combining multiple generative models (Gemini 2.5 Flash and Veo 3.1) with external services, enabling complex multi-step workflows (text analysis, prompt generation, safety validation, feedback generation, and media creation) to be consolidated into a single, scalable HTTP service.

🚀 What's next for C6 Fun Writing - English Writing to Creative Visuals

1.  Social Media Sharing & Community: Implement a secure, moderated social function allowing students to easily share their best AI-generated visual media (images/videos) and stories to social platforms and within the C6 community portfolio.

2.  Age-Group Curriculum Expansion: Utilize the PromptAgent to create a deeper, standardized curriculum aligned with the six age groups, ensuring a progressive increase in writing complexity.

Built With

Share this project:

Updates