ComfyUI Node
MOSS-TTS Load Model
Load MOSS-TTS Local Transformer v1.5 and its 48 kHz stereo audio tokenizer with ComfyUI/AIMDO memory tracking.
MOSS-TTS Load Model
- mosstts_model
◄modelMOSS-TTS Local Transformer v1.5 BF16 - OpenMOSS-Team (auto-download)►
◄dtypeauto►
◄attentionauto►
◄download_if_missingtrue►
CategoryMOSS-TTS
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | MOSS-TTS Local Transformer v1.5 BF16 - OpenMOSS-Team (auto-download) | Cataloged OpenMOSS MOSS-TTS Local Transformer v1.5. Main weights are stored in ComfyUI/models/mosstts/moss-tts-local-transformer-v1.5/. |
| dtype | COMBO | auto | auto uses the dtype declared by the bundled model config. Manual options are bf16 and fp16. |
| attention | COMBO | auto | auto uses FlashAttention 2 when compatible, SDPA on CUDA fallback, and eager on CPU. Manual eager forces the main model's plain attention path; the codec uses SDPA. |
| download_if_missing | BOOLEAN | true | Download missing main model and required audio-tokenizer model files. Disable for offline operation. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| mosstts_model | MOSSTTS_MODEL | — |