ComfyUI Node
Scenema Models Loader ⚡
Load the Scenema Audio components (audio DiT, Gemma-3 text encoder with GGUF support, audio VAE) as native comfy objects.
Scenema Models Loader ⚡
- model
- clip
- vae
◄transformer_name▾►
◄text_encoder_name▾►
◄pipeline_name▾►
◄vae_encoder_name▾►
◄keep_loadedtrue►
Category🤖 CCTech/Scenema
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| transformer_name | COMBO | scenema-audio-transformer-int8.safetensors (or the bf16 checkpoint, or a GGUF quant of either) from models/diffusion_models (unet). INT8 and GGUF weights remain quantized and dequantize one layer at a time. | |
| text_encoder_name | COMBO | Gemma-3 12B text encoder, .gguf (stays quantized) or safetensors, from models/text_encoders (clip). | |
| pipeline_name | COMBO | scenema-audio-pipeline.safetensors (or -pipeline-audio) from models/vae. Carries the text projection, embeddings connectors, audio VAE decoder and vocoder. | |
| vae_encoder_name | COMBO | scenema-audio-vae-encoder.safetensors from models/vae. Optional: needed to encode voice references only when the pipeline file above ships without its encoder (the full pipeline checkpoint already includes one). | |
| keep_loaded | BOOLEAN | true | Reuse the last-built model/clip/vae when these four filenames are unchanged, instead of rebuilding from disk every run (the DiT alone is 7-10 GB, the text encoder ~24 GB). Turn off to always force a fresh rebuild, e.g. after replacing a file on disk without renaming it. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| model | MODEL | — |
| clip | CLIP | — |
| vae | VAE | — |