ComfyUI Node
LTX 2.3 Unified Sampler (CRT)
A ComfyUI node in CRT/LTX2.3 with 27 inputs and 2 outputs.
LTX 2.3 Unified Sampler (CRT)
- models_pipe
- config_pipe
- images
- audio
◄workflow_modeI2V►
◄hqfalse►
◄live_previewfalse►
◄frame_count_from_audiofalse►
◄vae_decode_tiledfalse►
◄unload_model_before_vae_decodetrue►
◄low_vramtrue►
◄megapixels_target1.0►
◄aspect_ratio16:9 (Landscape)►
◄frame_count161►
◄v2v_modeDepth Control►
◄v2v_guide_strength0.60►
◄depth_megapixels0.30►
◄v2v_aspect_ratio16:9 (Landscape)►
◄sampler_maineuler►
◄sampler_refineeuler►
◄steps9►
◄generated_audio_gain_db-12.0►
◄firstframe_strength1.00►
◄depth_mouth_maskfalse►
◄mouth_detect_megapixels0.25►
◄mouth_single_itemtrue►
◄mouth_detect_chunk_size10►
◄mouth_mask_expand8►
◄mouth_mask_blur8.0►
CategoryCRT/LTX2.3
Inputs (27)
| Name | Type | Default | Description |
|---|---|---|---|
| models_pipe | LTX23_US_MODELS_PIPE | — | |
| config_pipe | LTX23_US_CONFIG_PIPE | — | |
| workflow_mode | COMBO | I2V | 3 options: I2V, T2V, V2V |
| hq | BOOLEAN | false | — |
| live_preview | BOOLEAN | false | — |
| frame_count_from_audio | BOOLEAN | false | — |
| vae_decode_tiled | BOOLEAN | false | — |
| unload_model_before_vae_decode | BOOLEAN | true | — |
| low_vram | BOOLEAN | true | When enabled, unloads the text encoder after conditioning and the VAE before sampling to reduce VRAM usage. The VAE is reloaded before decoding. |
| megapixels_target | FLOAT | 1.00.1–16 | — |
| aspect_ratio | COMBO | 16:9 (Landscape) | 20 options: 1:1 (Square), 2:3 (Portrait), 3:4 (Portrait), 4:5 (Portrait), 5:7 (Portrait), 5:8 (Portrait), +14 |
| frame_count | INT | 1611–4096 | — |
| v2v_mode | COMBO | Depth Control | 3 options: Depth Control, Outpaint, Upscale |
| v2v_guide_strength | FLOAT | 0.600–1 | — |
| depth_megapixels | FLOAT | 0.300.05–8 | — |
| v2v_aspect_ratio | COMBO | 16:9 (Landscape) | 20 options: 1:1 (Square), 2:3 (Portrait), 3:4 (Portrait), 4:5 (Portrait), 5:7 (Portrait), 5:8 (Portrait), +14 |
| sampler_main | COMBO | euler | 44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38 |
| sampler_refine | COMBO | euler | 44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38 |
| steps | INT | 94–128 | — |
| generated_audio_gain_db | FLOAT | -12.0-60–24 | — |
| firstframe_strength | FLOAT | 1.000–1 | — |
| depth_mouth_mask | BOOLEAN | false | — |
| mouth_detect_megapixels | FLOAT | 0.250–8 | Deprecated compatibility input. Ignored; mouth detection uses hardcoded fast mode. |
| mouth_single_item | BOOLEAN | true | Deprecated compatibility input. Ignored; hardcoded internally. |
| mouth_detect_chunk_size | INT | 100–4096 | Deprecated compatibility input. Ignored; hardcoded internally. |
| mouth_mask_expand | INT | 80–100 | Expands the mouth mask before applying it to the depth guide so lip motion influence covers a larger area. |
| mouth_mask_blur | FLOAT | 8.00–100 | Softens the mouth mask edges to avoid harsh transitions in the depth guide. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| audio | AUDIO | — |