The Transformation Engine

The Transformation Engine

LLM-powered structured prompts and transformations and other random wackiness for video generation models

Features

• Multi-Provider Architecture: Use any LLM (OpenRouter 100+ models, OpenAI, Gemini, local servers) for any task
• Intermediate-First Generation: Create model-agnostic semantic prompts, instantly export to Sora 2, Veo 3, or Generic formats
• Task-Level Granularity: Assign different providers/models per task (generate, mix, transform, etc.)
• Streaming Support: Live token streaming with tok/s metrics, cancel generation mid-stream
• Multi-Turn Conversations: Refine outputs iteratively with conversation tracking
• Token Tracking: Session-based usage dashboard with provider/task breakdown, export CSV/JSON
• Generate: Natural language → structured output (YAML, JSON, XML, Markdown, Natural Language)
• Transform: Convert between formats with mix options (reverse, compress, expand, technical, custom)
• Mix: Blend multiple prompts into hybrids
• Library: Local prompt management with search, favorites, versions
• Media: Image/video conditioning via vision API
• Transparency: Preview and edit all prompts before sending, customize system prompts, view intermediate representations
• Settings: 7 tabs (Providers, Tasks, API Key, Model, System Prompts, Tokens, Data)
• Privacy: 100% client-side, encrypted API keys per provider, network monitoring, audit logs

Screenshots

Main Interface

Three-panel layout with prompt library, input controls, and structured output.

Prompt Preview & Transparency

View and edit system + user prompts before sending to the API.

Media Conditioning & Mixing

Upload images/videos, mix multiple p