Trellis Text

From the project README

Native text-to-3D extension for Modly using the official TRELLIS text pipeline.

This extension is intentionally separate from the full TRELLIS.2 image/texturing extension. It is designed for users who only need prompt-to-mesh generation and do not want the additional image-to-mesh and mesh-texturing dependency surface.

Features Three Modly nodes backed by official TRELLIS text checkpoints: text-to-mesh-base — Base / low VRAM text-to-mesh-large — Large / balanced text-to-mesh — XL / original high-quality node kept for workflow compatibility Generator class: TrellisTextGenerator Official native TRELLIS text pipeline: TrellisTextTo3DPipeline GLB export through trellis.utils.postprocessingutils.toglb Localized auxiliary model references for TRELLIS text pipeline assets Reduced dependency surface compared with the full TRELLIS.2 extension

The runtime is text-to-mesh, but upstream Modly currently executes model workflow nodes through the image generation path and calls readFileBase64() before invoking the extension. For compatibility, every TRELLIS text node is therefore declared as image -> mesh in manifest.json.

The image input is a placeholder and is ignored by TrellisTextGenerator; generation is controlled by the Prompt parameter. In workflows, connect any valid image input or select an image before running, then enter the actual text prompt in the node parameters.