open-generative-ai-rs is a pure Rust orchestration gateway and web studio for chaining over 400 image, video, audio, and 3D generative AI models. It unifies local engines (sd.cpp, MLX, Wan2GP) and cloud providers behind a single type-safe Rust API with visual DAG workflow execution.
$ cargo install open-generative-ai-rs
This graphical studio is embedded directly inside the compiled binary and served via Axum at /studio.
open-genai generate image --model flux-dev --prompt "..."
Full endpoint schemas, multi-aspect-ratio scaling, and local inference flags.
Black Forest Labs 12B flow-matching transformer with photorealism and typographic fidelity.
Physics-accurate fluid dynamics, 1080p high bitrate rendering, and cinematic camera panning.
OpenAI video architecture with persistent world continuity and synchronized sound generation.
Open-weights 14B video foundation model. Executable via local Wan2GP Gradio server or cloud gateway.
Speech-driven portrait animator synthesizing natural micro-expressions, blinks, and head tilt.
Up to 14 multi-image inputs for strict character consistency, style transfer, and virtual try-on.
Comparing runtime efficiency, memory footprint, and deployment simplicity.
| Metric | Original (Next.js / Node) | open-generative-ai-rs | Performance Win |
|---|---|---|---|
| Binary Footprint | ~550 MB (node_modules) | 18 MB static binary | 30× smaller |
| Cold Start Time | 2.5s – 4.0s (V8 init) | < 15 ms | 200× faster |
| Idle Memory (RAM) | 180 – 250 MB | 11 MB | 20× less RAM |
| Concurrency Architecture | Single-threaded event loop | Tokio work-stealing pool | True CPU parallelism |
| Workflow Execution | Client-side ReactFlow loop | Native Rust DAG with Kahn's cycle check | Deterministic & headless |
| Offline Test Mode | Paid API calls required | Built-in --mock provider | Zero-cost CI/CD |
Integrate Open Generative AI directly into Python, Go, Node, or shell scripts.
Submit a text-to-image or image-to-image task and receive the completed asset URL.
curl -X POST http://127.0.0.1:8080/api/v1/generate/image \
-H "Content-Type: application/json" \
-d '{"model": "flux-dev", "prompt": "cyberpunk skyline", "aspect_ratio": "16:9"}'
Trigger a multi-step DAG pipeline (e.g. AI Influencer, Popcorn Storyboard) with variable passing.
curl -X POST http://127.0.0.1:8080/api/v1/workflows/run \
-H "Content-Type: application/json" \
-d '{"template_id": "ai-influencer", "inputs": {"prompt": "news anchor"}}'