ComfyUI Node
AuK Whisper Transcribe X
Transcribe audio to craft edit instructions or check AuK output. Uses local models under models/whisper or models/audio_encoders; can optionally fetch a known Whisper checkpoint once, explicitly enabled.
AuK Whisper Transcribe X
- audio
- STRING
◄model▾►
◄download_if_missingfalse►
◄model_sizelarge-v3►
◄language▾►
◄task▾►
CategoryWorkflowX/Audio/AuK
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| audio | AUDIO | Audio to transcribe - AuK's output as a quality check, or source audio before writing a replace/extract instruction. | |
| model | COMBO | Whisper checkpoint folder from models/whisper or models/audio_encoders. Names without a local folder download only when download_if_missing is on. | |
| download_if_missing | BOOLEAN | false | When the selected model is not on disk, download it once from Hugging Face into models/whisper/ and reuse the local copy afterwards. Off keeps the node fully offline. |
| model_size | COMBO | large-v3 | Fallback size to fetch when download_if_missing is on and the selected model is not itself a known download size. Saved under models/whisper/<size>/. |
| language | COMBO | Spoken language hint; auto detects it. | |
| task | COMBO | transcribe keeps the spoken language; translate outputs English. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| STRING | STRING | — |