Skip to content

CHANGELOG.md

Latest commit

 

History

History
70 lines (56 loc) · 2.72 KB

CHANGELOG.md

File metadata and controls

70 lines (56 loc) · 2.72 KB
Raw

Changelog

All notable changes to OpenFrame AI will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.0.0] — 2026-02-23

Added

Multi-Agent AI Pipeline

  • 7 specialized AI agents: Creative Director, Brand Stylist, Product Stylist, Casting Scout, Cinematographer, Director, Sound Designer
  • Creative Director self-critique and revision loop (agent 1 → critic → revise)
  • Fan-out parallelism: Brand Stylist triggers Product Stylist, Casting Scout, and Cinematographer simultaneously
  • Fan-in synchronization: Director waits for all 3 parallel agents before generating shot list
  • Per-agent LLM model selection (27 models across 7 providers via OpenRouter)
  • Per-agent temperature control
  • 5 ad-type presets: Fashion & Luxury, Commercial, Beauty, UGC, Cinematic Brand Film

Visual Canvas

  • React Flow-based visual canvas for ad composition
  • Asset nodes: character, environment, product, camera, voiceover, music, image, video
  • Scene nodes: start/end frames with image generation, video generation, and audio
  • Smart wire connections between assets and scenes
  • Drag-and-drop canvas layout

Image Generation

  • 4 image providers: Seedream, Nano Banana, GPT Image, Z Image (via kie.ai)
  • Per-asset and per-frame model selection with quality controls
  • Reference image support (connected assets auto-included)
  • Image upload and history navigation
  • Aspect ratio controls

Video Generation

  • 15+ video models including Kling 3.0, Wan 2.1, Veo 3, Seedance 2
  • Three audio modes per scene: silent, audio-native, talking-head
  • Per-slot audio mode (separate start/end video audio)
  • InfiniTalk integration for talking-head videos
  • Video duration and resolution controls

Audio & Voice

  • ElevenLabs text-to-speech for voiceovers
  • Voice picker with stability, similarity, style, and speed controls
  • AI music generation (Suno/Udio via kie.ai)
  • Per-scene dialogue with speaker assignment
  • SRT subtitle generation via Whisper

Project Management

  • Project list with search, folders, and drag-to-organize
  • 3D animated folder cards
  • Project save/load with full state persistence
  • Manual mode (build from scratch without AI agents)
  • JSON import/export
  • Project rename and delete

Infrastructure

  • FastAPI backend with SSE real-time streaming
  • SQLite database for API logs and auth
  • Runtime API key management (set keys via UI, no restart needed)
  • Docker Compose one-command deployment
  • Cross-platform start scripts (PowerShell + Bash)
  • User authentication (signup, signin, password reset)
  • API cost tracking across all providers

For older development history, see git log.