ComfyUI Node
ID-LoRA Two-Stage Sampler
Generate audio+video with speaker identity transfer using the ID-LoRA two-stage pipeline. Stage 1 generates at the specified resolution, stage 2 refines at 2x resolution.
ID-LoRA Two-Stage Sampler
- pipeline
- conditioning
- first_frame
- reference_audio
- Video
◄seed42►
◄height512►
◄width512►
◄num_frames121►
◄num_inference_steps30►
◄frame_rate25.0►
◄video_guidance_scale3.0►
◄audio_guidance_scale7.0►
◄auto_resolutiontrue►
◄max_resolution512►
◄hq_modetrue►
CategoryID-LoRA
Inputs (15)
| Name | Type | Default | Description |
|---|---|---|---|
| pipeline | ID_LORA_PIPELINE | Loaded ID-LoRA two-stage pipeline. | |
| conditioning | ID_LORA_CONDITIONING | Encoded prompt conditioning. | |
| seed | INT | 420–2147483647 | — |
| height | INT | 51264–2048 | Stage 1 height. Output will be 2x this value. |
| width | INT | 51264–2048 | Stage 1 width. Output will be 2x this value. |
| num_frames | INT | 1211–1000 | — |
| num_inference_steps | INT | 301–200 | Number of denoising steps for stage 1. Stage 2 uses 3 fixed steps. |
| frame_rate | FLOAT | 25.01–120 | — |
| video_guidance_scale | FLOAT | 3.00–30 | — |
| audio_guidance_scale | FLOAT | 7.00–30 | — |
| auto_resolution | BOOLEAN | true | Auto-detect resolution from first-frame aspect ratio. |
| max_resolution | INT | 51264–2048 | Maximum long-side resolution for auto-resolution (ignored when auto_resolution is off). |
| hq_mode | BOOLEAN | true | Use res2s second-order sampler for higher quality (slower). |
| first_frameopt | IMAGE | Optional first-frame image for face conditioning. | |
| reference_audioopt | AUDIO | Optional reference audio for speaker identity transfer. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Video | VIDEO | Generated video with audio (2x input resolution). |