LTX 2.3 Workflow Collection, including Wan 2.2 and HuMo refinement
About this model
LTX Video 2.3 — Multi-Stage Workflow Collection
A collection of LTX Video 2.3 workflows built for high-end hardware, focused on quality over speed. Designed without get/set nodes, AIO functions aside from text/image combined workflows or subgraphs.
Pipeline Variants
Two-Stage
Draft → Upsample → RefineStandard pipeline
Triple-Stage
Quarter → Half → FullUp to 4x faster · output must be div-by-128
HuMo
Two-stage + motion pass - Audio-driven human motion conditioning
Wan 2.2 LN
Two-stage + refinement pass - lower denoise required than HuMo
Features
Audio latent retention — first-stage audio latent is carried into refinement rather than re-encoded, preserving timing and texture coherence across passes
Per-stage sampler & scheduler tuning — each stage uses independently configured samplers and schedulers
RTX Video Super Resolution output — easily removed if not needed
Dimension & frame calculator nodes — can be swapped for any input source; note that refinement stages have strict dimensionality requirements
Workflow Types
T2V / I2V toggle · First & Last Frame · Audio I2V · Identity LoRA · Video Continuation
Most types are available across all three pipeline variants.
The active repo is found at I post here as direction to the full collection. Please see the readme there for more details.
Tags
Related Models
Similar AI models you may like
ON-THE-FLY 实时生成!Wan-AI 万相/ Wan2.1 Video Model (multi-specs) - CausVid&Comfy&Kijai - workflow included
【WAN2.1】IMG to VIDEO
ComfyUI Image Workflows
WAN 2.2 Workflow T2V-I2V-T2I (Kijai Wrapper)
Hunyuan 🌻 AllInOne
Moody Simple Zimage Turbo/Distilled Workflow