Skip to content

README.md

Latest commit

 

History

History
59 lines (44 loc) · 2.68 KB

README.md

File metadata and controls

59 lines (44 loc) · 2.68 KB
Raw

AI Transcription & Classifier

Transcribes audio and video using Whisper (via DeepInfra or Replicate) and classifies the output using Gemini 1.5 & 3 (via Vertex AI). This app is optimized for Vercel, featuring high-speed transcription, intelligent classification, and persistent storage.

Key Features

  • Multi-Model Selection: Choose between DeepInfra (Whisper) or Replicate (Insanely Fast Whisper) for transcription.
  • Gemini 3 Support: Integrated with Gemini 3 Flash and 3.1 Pro via Vertex AI.
  • Implicit Caching: 30-50% faster classification and 90% lower prompt costs by leveraging Vertex AI's automatic caching of repeated system instructions.
  • Direct-to-Blob Upload: Bypasses Vercel's 4.5MB serverless limit, supporting files up to 50MB+.
  • Duplicate Upload Prevention: Automatically detects if a file already exists in cloud storage by name, skipping redundant uploads to save time and bandwidth.
  • High-Concurrency Engine: Parallel processing for up to 20 files simultaneously, with automatic background overflow for larger batches.
  • Real-time Batch Status: Individual tracking of upload/processing status for every file in a batch with selective result viewing.
  • Lazy Audio Navigation: Optimized search page that lazily loads audio segments only when played, preventing network congestion in large result lists.

Setup

  1. Clone & install

    npm install

  2. Set environment variables — create a .env file at the project root:

    # Transcription
DEEPINFRA_API_KEY=your_deepinfra_key_here
REPLICATE_API_TOKEN=your_replicate_token_here

# Vertex AI (Classification)
GOOGLE_CLOUD_PROJECT=your_project_id
GOOGLE_CLOUD_CLIENT_EMAIL=your_service_account_email
GOOGLE_CLOUD_PRIVATE_KEY="your_private_key_content"

# Storage
BLOB_READ_WRITE_TOKEN=your_vercel_blob_token_here

  3. Run locally

    npm run dev

    Open http://localhost:3000

Usage

  1. Upload New: Select any audio/video file and choose your preferred Transcription and Classification models.
  2. Browse Library: Click the library tab to see existing files in your cloud storage and process them instantly.
  3. Dashboard: View timing comparisons between models and access historical classification results.
  4. Background Processing: Refer to AZURE_DURABLE_FUNCTIONS.md for details on setting up automated, long-running transcription tasks.

Deploy to Vercel

vercel deploy

Ensure all your .env variables are configured in the Vercel dashboard.

Optimized with Vertex AI Context Caching and Next.js Server Actions.