ComfyUI Node
ThinkingLLM Gemma 4 Audio (GGUF)
A ComfyUI node in ThinkingLLM with 11 inputs and 2 outputs.
ThinkingLLM Gemma 4 Audio (GGUF)
- audio
- RESPONSE
- RAW_TRACE
◄model_namegemma-4-12b-it-BF16.gguf [~24GB]►
◄custom_promptAudio analysis: transcribe the speech in the original language, then summarize the important points. If there is no clear speech, describe the audible scene and relevant sounds.►
◄max_tokens2048►
◄keep_model_loadedfalse►
◄seed1►
◄stream_tokens_to_terminalfalse►
◄enable_thinkingfalse►
◄auto_finalization_retryfalse►
◄hf_token►
◄audio_file_path►
CategoryThinkingLLM
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | gemma-4-12b-it-BF16.gguf [~24GB] | Gemma 4 audio-capable GGUF model. Only Gemma 4 E2B, E4B, and 12B are listed; 26B/31B variants are image/text-only for this purpose. |
| custom_prompt | STRING | Audio analysis: transcribe the speech in the original language, then summarize the important points. If there is no clear speech, describe the audible scene and relevant sounds. | Audio instruction sent with the AUDIO input. Gemma 4 audio works best with short 16 kHz mono WAV-style input; Comfy AUDIO is converted to WAV before inference. |
| max_tokens | INT | 204864–32768 | Maximum new tokens to generate. Larger values give more room for reasoning but increase runtime and memory use. |
| keep_model_loaded | BOOLEAN | false | Keep the GGUF model in RAM/VRAM after the run so repeated prompts skip model loading. Disable if you need memory back for other nodes. |
| seed | INT | 11–4294967295 | Sampling seed. The node also uses fixed-seed prompt persistence, so identical inputs can reuse the saved result. |
| stream_tokens_to_terminal | BOOLEAN | false | Print generated tokens live in the ComfyUI terminal. Useful for long runs and backend troubleshooting. |
| enable_thinking | BOOLEAN | false | Gemma 4 can reason, but audio transcription and short analysis are usually clearer with thinking disabled. |
| auto_finalization_retry | BOOLEAN | false | If enabled, runs an extra LLM completion when the first output is empty or reasoning-only. Disabled by default so one node execution performs one generation pass. |
| hf_token | STRING | Optional Hugging Face access token for private or gated GGUF/mmproj downloads. It is passed only to the download call, never logged or cached, and the in-memory copy is dropped after the download attempt. Clear this field before saving or sharing workflows. | |
| audioopt | AUDIO | — | |
| audio_file_pathopt | STRING | Optional local audio file path. M4A, MP3, WAV, FLAC, and other FFmpeg-readable files are decoded to 16 kHz mono WAV before inference. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| RESPONSE | STRING | — |
| RAW_TRACE | STRING | — |