My ComfyUI workflows (UI format). To use one, drag the .json onto the ComfyUI canvas or copy it into ComfyUI/user/default/workflows/.
| Workflow | Description |
|---|---|
LTX-2.5_FLF_V2V_ICLoRA_Native_4K_8K.json |
LTX-2.5 video-to-video with first/last-frame guides, IC-LoRA and Tiled Fusion, built on the Native 4K/8K structure (target: RTX 3090, 24 GB) |
LTX-2.5_FLF_V2V_ICLoRA_SDR_to_HDR_Native_4K_8K.json |
Same FLF + IC-LoRA stage 1, then Lightricks' SDR-to-HDR Tiled Fusion pipeline: HDR-A (SDR-to-HDR IC-LoRA, FullHD) and HDR-B (refine-details IC-LoRA, 4K/8K, guided by the HDR-A result, as in Lightricks' HDR-to-HDR graph). Source through yRead; the decode is chained into ACEScct -> linear ACEScg, into yWrite (EXR) and yViewer (no video at the end) (read the EXRs as ACEScg in Nuke, not ACEScct) |
LTX-2.5_FLF_V2V_ICLoRA_SDR_to_HDR_Native_4K_8K_Subgraphs.json |
Same as the SDR-to-HDR workflow, tidied into subgraphs: Load Models, Prompt, Source Video + Sizes, Stage 1, SDR to HDR, Upscale, HDR Refine, Decode HDR |
LTX-2.5_FLF_V2V_ICLoRA_TiledFusion.json |
Earlier single-pass Tiled Fusion version of the above |
*_GGUF_Q6K.json (one per LTX workflow above) |
Same workflows with the diffusion model loaded as GGUF Q6_K through Unet Loader (GGUF) (ComfyUI-GGUF); the text encoder stays the Gemma comfy int8. Download links are in the "Downloads (GGUF Q6_K)" note in each workflow. |
image_flux3_klein_distilled_image_edit.json |
FLUX Klein distilled image edit |
Viewer_test.json |
Viewer test |
The SDR-to-HDR workflows use the studio nodes yRead / yWrite (vfx_io) and yViewer (vfx_viewer). Their 4K/8K pass also needs ltx-2.5-22b-ic-lora-refine-details-1.0.safetensors in models/loras/ (Lightricks/LTX-2.5-22b-IC-LoRA-Refine-Details).