ComfyUI Node
LTX Infinite Director Sampler V7 (Ref-Frame Lock)
Reference-frame latent extension for LTX 2.3. The last carry_frames pixel frames from chunk N are VAE-encoded and locked (noise_mask=0) at the start of chunk N+1's latent. The model generates only the new region in a single pass, anchored to real prior content — same architecture as Kijai NativeLooping and LTX extend workflows.
LTX Infinite Director Sampler V7 (Ref-Frame Lock)
- model
- clip
- audio_vae
- vae
- spatial_upscaler
- bg_audio
- last_chunk_frames
- audio
- stage1_frames
◄start_second0.00►
◄end_second85.00►
◄duration_seconds85.00►
◄start_frame0►
◄end_frame2040►
◄duration_frames2040►
◄timeline_data{}►
◄local_prompts►
◄segment_lengths►
◄global_prompt►
◄guide_strength►
◄epsilon0.0010►
◄frame_rate24.00►
◄display_modeseconds►
◄custom_width720►
◄custom_height1280►
◄resize_methodmaintain aspect ratio►
◄divisible_by32►
◄img_compression0►
◄generate_audiotrue►
◄custom_audio_onfalse►
◄motion_guide_ontrue►
◄chunk_duration_seconds17.0►
◄auto_chunk_threshold20.0►
◄carry_frames73►
◄carry_strength1.00►
◄crossfade_frames0►
◄ic_lora_nameNone►
◄ic_lora_strength1.00►
◄stage1_steps8►
◄stage2_steps4►
◄stage2_denoise0.42►
◄cfg1.0►
◄seed1000►
◄filename_prefixmuse►
◄bg_volume1.00►
CategoryLTXInfiniteDirector
Inputs (42)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| clip | CLIP | — | |
| audio_vae | VAE | — | |
| start_second | FLOAT | 0.000–3600 | — |
| end_second | FLOAT | 85.000–3600 | Total video length in seconds. |
| duration_seconds | FLOAT | 85.000–3600 | — |
| start_frame | INT | 00–86400 | — |
| end_frame | INT | 20400–86400 | — |
| duration_frames | INT | 20400–86400 | — |
| timeline_data | STRING | {} | — |
| local_prompts | STRING | — | |
| segment_lengths | STRING | — | |
| global_prompt | STRING | — | |
| guide_strength | STRING | — | |
| epsilon | FLOAT | 0.00100–1 | — |
| frame_rate | FLOAT | 24.001–120 | — |
| display_mode | COMBO | seconds | 2 options: seconds, frames |
| custom_width | INT | 72064–4096 | — |
| custom_height | INT | 128064–4096 | — |
| resize_method | COMBO | maintain aspect ratio | 4 options: maintain aspect ratio, stretch to fit, crop, pad |
| divisible_by | INT | 321–256 | — |
| img_compression | INT | 00–51 | — |
| generate_audio | BOOLEAN | true | LTX generates audio from [SOUNDS] prompts. Can be combined with BG Audio. |
| custom_audio_on | BOOLEAN | false | Use audio file from the timeline AUDIO track. Can be combined with Generate Audio. |
| motion_guide_on | BOOLEAN | true | Use motion guide segments from the timeline. |
| vae | VAE | — | |
| spatial_upscaler | LATENT_UPSCALE_MODEL | — | |
| chunk_duration_seconds | FLOAT | 17.02–120 | — |
| auto_chunk_threshold | FLOAT | 20.00–3600 | — |
| carry_frames | INT | 731–240 | Reference frames from previous chunk to lock at the start of each new chunk. More = stronger continuity, bigger latent. 73 ≈ 3s at 24fps. |
| carry_strength | FLOAT | 1.000–1 | Unused in V4 (kept for workflow compatibility). |
| crossfade_frames | INT | 00–120 | — |
| ic_lora_name | COMBO | None | 1 options: None |
| ic_lora_strength | FLOAT | 1.00-10–10 | — |
| stage1_steps | INT | 81–50 | — |
| stage2_steps | INT | 41–50 | — |
| stage2_denoise | FLOAT | 0.420–1 | — |
| cfg | FLOAT | 1.00–20 | — |
| seed | INT | 10000–18446744073709550000 | — |
| filename_prefix | STRING | muse | — |
| bg_volume | FLOAT | 1.000–2 | — |
| bg_audioopt | AUDIO | — |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| last_chunk_frames | IMAGE | — |
| audio | AUDIO | — |
| stage1_frames | IMAGE | — |