SVF vs SVF2: What It Means for Performance and Your APS Pipeline

If you’re building on Autodesk Platform Services (APS, formerly Forge) and using the Model Derivative API to translate CAD/BIM files for viewing, you’ve encountered two output format options: SVF and SVF2. Which one you choose has a measurable impact on load performance, streaming behavior, and pipeline architecture—especially at scale.

What SVF and SVF2 Actually Are

SVF (Simple Viewer Format) is Autodesk’s original web-viewable format. When you translate a RVT, DWG, or other CAD file via the Model Derivative API, SVF produces a set of files that the Viewer SDK loads to render the model in a browser.

SVF2 is the second-generation format introduced to address SVF’s performance limitations at scale. It restructures the same geometry and metadata into a format optimized for streaming, sharing materials across a scene, and enabling progressive loading of large models.

Key Differences

Dimension	SVF	SVF2
Output file count	Many small files	Fewer, larger files
Loading strategy	Download all then render	Progressive streaming
Material sharing	Per-model	Cross-model (shared pack)
Large model support	Struggles at scale	Designed for large scenes
Time to first render	Slower	Faster (streaming)
Viewer SDK support	v7+	v7+ (full from v7.50)
Storage size	Larger total	~30-50% smaller

What This Means for Your Pipeline

1. Translation Job Setup

To request SVF2, you specify it in the Model Derivative API job payload:

{
  "input": { "urn": "dXJuOmFkc2s..." },
  "output": {
    "formats": [
      {
        "type": "svf2",
        "views": ["2d", "3d"]
      }
    ]
  }
}

The job outputs go to your OSS bucket. SVF2 jobs typically take slightly longer to complete than SVF for the same model because of the additional packing step.

2. Viewer Initialization

Both formats load via the same Viewer SDK loadDocumentNode API. The viewer auto-detects the format from the manifest. No code change is needed on the viewer side when switching from SVF to SVF2—just ensure you’re on Viewer SDK v7.50 or later for full SVF2 support.

3. Metadata Extraction—No Difference

Whether you use SVF or SVF2 as output, the metadata extraction endpoints (object tree, properties, materials) behave identically. The format choice is purely a rendering/streaming optimization. If your primary use case is metadata extraction for downstream AI/ERP workflows, the format choice matters less—though SVF2 is still recommended for consistency if you’re also rendering.

Which Should You Use?

The Pipeline Architecture Implication

For teams building AI-enabled APS pipelines—where extracted metadata feeds into LLMs, quoting systems, or inspection agents—SVF2 reduces your infrastructure cost (fewer bytes in/out of OSS) and improves the user experience when engineers need to view drawings alongside AI outputs. Treat the format choice as an infrastructure decision you make once, early, and enforce consistently across your translation pipeline.