Routing Prompts to Video Engines with Orchestration Models
Behind the scenes, Pipevideo acts as an intelligent router between your prompts and the most appropriate video generation infrastructure. This post explains how our routing system works and how to leverage it for your use cases.
The Two-Layer Routing System
Pipevideo uses a two-layer approach to route requests:
Layer 1: Engine Selection
The engine parameter determines the visual style and output format:
- HyperFrames: Generates high-quality video frames with detailed motion control
- Lottie: Produces vector-based animations perfect for web and mobile UI
Each engine has different strengths:
| Engine | Best For | Output Format |
|---|---|---|
| HyperFrames | Product demos, realistic motion | MP4, high resolution |
| Lottie | UI animations, icons, loading states | JSON (native Lottie) |
Layer 2: Orchestration Model Selection
The model parameter determines how your prompt is interpreted and executed:
- Claude Opus 4.8: Complex prompts requiring precise visual reasoning
- Kimi K2.6: Balanced performance for general video generation
- GPT-5.5: Strong prompt adherence and detailed generation
- Gemini 3.5 Flash: Fast generation for prototyping
How Routing Works
When you send a request, here's what happens:
- Prompt Analysis: The orchestration model analyzes your prompt for video generation requirements
- Engine-Specific Translation: The model translates your natural language into engine-specific parameters
- Generation Execution: The engine executes the generation with the translated parameters
- Post-Processing: Video output is normalized and returned in a consistent format
Choosing the Right Combination
For Product Marketing Videos
{
"model": "anthropic/claude-opus-4.8",
"engine": "hyperframes",
"input": "A 15-second product showcase featuring our electric vehicle"
}Why this works: Claude Opus provides detailed visual reasoning for complex product shots, while HyperFrames delivers high-quality output.
For App UI Animations
{
"model": "moonshotai/kimi-k2.5",
"engine": "lottie",
"input": "A success checkmark animation with a subtle bounce effect"
}Why this works: Kimi K2.5 provides efficient processing for straightforward animations, and Lottie produces lightweight vector output perfect for apps.
Dynamic Routing (Coming Soon)
We're building automatic routing that selects the optimal engine/model combination based on your prompt content. This will:
- Analyze your prompt for complexity and style requirements
- Compare current provider pricing and latency
- Select the most cost-effective option that meets your quality needs
Model-Specific Pricing
Different models have different pricing based on their capabilities:
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Claude Opus 4.8 | $5.00 | $25.00 |
| Kimi K2.6 | $0.68 | $3.42 |
| GPT-5.5 | $5.00 | $30.00 |
| GPT-5.4 Nano | $0.20 | $1.25 |
See our Rankings page for live pricing and performance comparisons.
Advanced: Custom Routing Rules
Enterprise customers can define custom routing rules:
- Always use specific models for brand consistency
- Fallback providers if primary provider is down
- Latency thresholds to ensure timely responses
- Cost caps with automatic downgrading to cheaper models
Contact us to learn about enterprise routing features.
Have questions about routing? Check our documentation or join our Discord community.