MintTales - 10.3

• 

  Genre Selection

• 

  Home Page

• 

  Workflow

Inspiration:- Stories deserve more than just text on a screen. We wanted to build a platform where AI doesn't just write stories & it brings them to life with voice, visuals, and on-chain permanence. The idea of combining generative AI with blockchain felt like a natural fit: create something beautiful, then own it forever.

What it does:- MintTales is an end-to-end AI storytelling platform. You pick a genre, co-author a story with Gemini AI through a conversational chat, then with one click: ElevenLabs narrates it with a genre-matched emotional voice, Gemini generates cinematic scene illustrations, a network of 5 autonomous Fetch.ai agents analyzes your story's quality, virality, and NFT value and you can mint the whole thing as an NFT on Solana's devnet through your Phantom wallet. Everything is stored in MongoDB for the community gallery.

How we built it Next.js 15 + TypeScript + Tailwind CSS v4 - Full-stack app with App Router Google Gemini (gemini-2.5-flash) - Powers the story co-authoring chat, scene extraction, agent analysis, and image generation (gemini-2.5-flash-image) ElevenLabs - Voice narration with genre-specific voices and tuned emotional settings (stability, style, similarity) per genre Solana (Devnet) - Real on-chain minting via the Memo Program, with Phantom wallet signing and automatic devnet SOL airdrop MongoDB Atlas - Persistent story storage, gallery, and database for the ScoutAgent to analyze trends Fetch.ai - 5 autonomous agents (Scout, Critic, Audience, Director, Valuation) that collaborate in a decentralized pipeline with protocol messages and FET token rewards Framer Motion - Cinematic story player with Ken Burns effect, waveform visualizer, and animated agent network pentagon graph

Challenges we ran into:- Image generation reliability - Our initial image provider (Pollinations.ai) kept returning 429 rate limits. We pivoted to Gemini's native image generation model, which turned out to produce higher quality cinematic scenes. Voice-genre mismatch - All stories initially sounded the same regardless of genre. We solved this by mapping each genre to a specific ElevenLabs voice and tuning the voice settings (stability, style, similarity_boost) for emotional delivery. Solana transaction signing - Building a flow where the server constructs the transaction but the client signs it with Phantom required careful serialization/deserialization of transaction bytes. Syncing audio with scenes - Timing the Ken Burns scene transitions to match the narration audio required word-count-based duration calculations.

Accomplishments that we're proud of:- a)A fully working pipeline from prompt to minted NFT in under 3 minutes b)Real on-chain Solana transactions with Phantom wallet integration c)Genre-aware emotional narration horror stories sound ominous, comedies sound playful d)The Fetch.ai agent network visualization with a live pentagon graph, protocol messages, and FET token economics e)AI-generated cinematic scene illustrations that match the story's mood.

What we learned:- 1)How Solana's Memo Program works for storing metadata on-chain without needing a full NFT program. 2)ElevenLabs voice settings (stability, style, similarity_boost) dramatically change how a story feels. 3)Multi-agent architectures can provide genuinely useful layered analysis when each agent has a distinct role. 4)Gemini's multimodal capabilities extend beyond text & it can generate high-quality images directly.

What's next for MintTales:- 1)Upgrade to full Metaplex NFTs with cover art and audio metadata. 2)Add collaborative story editing where multiple users contribute chapters. 3)Integrate Veo for AI video generation from scenes. 4)Build a marketplace where users can trade story NFTs using the Fetch.ai valuation as a price guide. 5)Add more genres and let users create custom voice profiles

Built With

• elevenlabs
• fetch.ai
• framermotion
• googlegemini
• javascript
• mongodb
• nextjs
• react19
• solanadevnet
• typescript