ComfyUI Node
ACE-Step Captioner
A ComfyUI node in Audio/ACE-Step with 12 inputs and 3 outputs.
ACE-Step Captioner
- audio
- caption
- style_tags
- full_description
◄model_idACE-Step/acestep-captioner►
◄deviceauto►
◄dtypeauto►
◄custom_prompt*Task* Describe this audio in detail►
◄max_new_tokens1024►
◄temperature0.3►
◄top_p0.90►
◄top_k50►
◄repetition_penalty1.1►
◄seed0►
◄chunk_length_s30►
CategoryAudio/ACE-Step
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| audio | AUDIO | Input audio to caption/describe. | |
| model_id | COMBO | ACE-Step/acestep-captioner | Select the captioner model. Can be a local path or HuggingFace ID. |
| device | COMBO | auto | Inference device. Use 'auto' or 'mps' for Mac. |
| dtype | COMBO | auto | Model precision. 'auto' uses float16 for CUDA and float32 for CPU/MPS. |
| custom_promptopt | STRING | *Task* Describe this audio in detail | Custom prompt for captioning. Default is the recommended prompt from ACE-Step. |
| max_new_tokensopt | INT | 102464–4096 | Maximum number of tokens to generate. Increase for longer descriptions. |
| temperatureopt | FLOAT | 0.30–1 | Sampling temperature. Lower values (0.1-0.3) are more deterministic and accurate. |
| top_popt | FLOAT | 0.900–1 | Nucleus sampling: cumulative probability threshold. |
| top_kopt | INT | 500–1000 | Top-K sampling. 0 = disabled. Lower values can improve accuracy. |
| repetition_penaltyopt | FLOAT | 1.11–2 | Penalty for repeating tokens. Increase if output gets stuck in loops. |
| seedopt | INT | 00–4294967295 | Random seed for reproducible results. 0 for random. |
| chunk_length_sopt | FLOAT | 300–300 | Audio chunk length in seconds for processing long audio. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| caption | STRING | — |
| style_tags | STRING | — |
| full_description | STRING | — |