Brandframe

• 

1. PROJECT NAME

BrandFrame

---

2. TAGLINE / ONE-LINER

AI-native video platform that automatically places contextually relevant advertisements into uploaded videos using a 13-step computer vision pipeline — with immutable ad manifests locked on Backblaze B2.

---

3. DESCRIPTION

The Problem

Video creators and advertisers struggle with:

• Manual, time-intensive ad placement in video content
• Lack of verification that ads were actually displayed
• No tamper-proof record of ad placements
• Limited monetization options for organic video content

The Solution

BrandFrame is an AI-native video ad platform that automates in-scene advertising. Upload any video → our 13-step Python pipeline detects optimal ad placement opportunities → places contextually relevant AI-generated ads → locks the verification manifest on Backblaze B2 Object Lock for 365 days.

How It Works

Upload & Analyze

• User uploads video to BrandFrame
• Deepgram ASR transcribes all audio with speaker diarization
• Mistral Pixtral identifies optimal surface regions for ad placement (screens, walls, objects)
• Scene detection maps content context for relevance matching

AI Ad Generation

• Gemini 2.5 Flash generates contextual ad creatives for detected surfaces
• Image inpainting seamlessly blends ads into the video scene
• Mistral AI critic validates ad placement quality and context alignment
• Three ad layer types offered:
  • Chat Sponsored Cards (overlay)
  • Natural-Break Mid-Rolls (scene cuts)
  • In-Scene Pause Ads (before→after crossfade transitions)

Verification & Immutability

• Genblaze manifest generated with SHA-256 hashes of all ad placements
• Manifest locked on Backblaze B2 with Object Lock (365-day retention)
• Public verification page lets anyone audit ad authenticity

Tech Stack

• Frontend: Next.js 15 + TypeScript + Tailwind CSS + shadcn/ui
• Theme: Dark warm aesthetic (background hsl(20 20% 6%), accent orange hsl(18 88% 54%))
• Pixel-art identity throughout the UI
• Database: Drizzle ORM + @libsql/SQLite (Turso-ready)
• Video AI: Deepgram (ASR), Mistral Pixtral (surface detection), Mistral (critic), Gemini 2.5 Flash (image inpainting)
• Storage & Verification: Backblaze B2 (Object Lock), Genblaze (manifest generation with BGE-M3 embeddings)
• Search: Hybrid search + RAG with timestamp chips

Key Features

✅ 13-step automated ad placement pipeline ✅ Three ad layer types (chat card, mid-roll, in-scene pause) ✅ AI-generated contextual ad creatives ✅ Immutable verification manifests on B2 Object Lock ✅ Public /verify page for ad authenticity ✅ RAG-powered chat with timestamp citations ✅ 100% hackathon provider compliance (B2, Mistral, Gemini, Deepgram, Genblaze/BGE-M3)

Built By

Solo build by @codedbytahir — Backblaze Generative Media Hackathon 2026

---

4. DEMO VIDEO

[PASTE YOUTUBE/VIMEO URL HERE]

---

5. TRY IT OUT / LINKS

- GitHub: https://github.com/codedbytahir/brandframe

6. BUILT WITH

• Next.js 1
• TypeScript
• Tailwind CSS
• shadcn/ui
• Deepgram (ASR)
• Mistral AI (Pixtral + Critic)
• Gemini 2.5 Flash (Image Inpainting)
• Backblaze B2 (Object Lock)
• Genblaze + BGE-M3
• Drizzle ORM
• SQLite / Turso

---

7. HACKATHON CATEGORY

• Best Generative Media Hack
• Best Use of Backblaze B2

---

8. INNOVATION

BrandFrame reimagines video monetization through AI automation. By combining computer vision surface detection, generative image inpainting, and immutable blockchain-style verification manifests on B2 Object Lock, we create a trustless ad placement system. The 13-step pipeline from upload → analysis → ad generation → placement → verification is fully automated and auditable.

---

9. WHAT'S NEXT FOR BRANDFRAME

• Multi-language ad generation
• Real-time streaming ad insertion
• Advertiser dashboard with analytics
• Tokenized ad verification NFTs
• Integration with major video platforms

---

Generated for BrandFrame DevPost Submission — Backblaze Generative Media Hackathon 2026

Built With

• ai
• genblazesdk
• next
• tsx