LTX 2.3 3D Real workflow ( 3D Render To Realistic Video) — v1.0

This workflow is for turning 3D renders into realistic video: provide an image or video plus text conditioning, and it produces decoded content and an assembled video.

The graph loads and resizes media, preprocesses it, applies LTXV conditioning, samples, decodes content, and assembles video.

`LoadImage` and `VHS_LoadVideo` provide image or video inputs, while `CLIPTextEncode` provides text conditioning.

`VAEDecode` provides decoded content, while `CreateVideo` and `VHS_VideoCombine` assemble video.

Audio and video latent data pass through `LTXVAudioVAELoader`, `LTXVAudioVAEDecode`, `LTXVConcatAVLatent`, and `LTXVSeparateAVLatent`.

Use the workflow file `ltx233DRealWorkflow3D_v10.json`; the listed release is `v1.0`.

The workflow requires these node packs: `ComfyMath`, `comfyui_essentials`, `comfyui-depthcrafter-nodes`, `comfyui-kjnodes`, `ComfyUI-LTXVideo`, `comfyui-videohelpersuite`, and `RES4LYF`.

The workflow requires `gemma_3_12B_it.safetensors` (text encoder), `ltx-2-spatial-upscaler-x2-1.0.safetensors` (upscaler), `ltx-2.3-22b-dev-fp8.safetensors` (checkpoint), `ltx-2.3-22b-dev.safetensors` (VAE), `ltx-2.3-22b-distilled-lora-384.safetensors` (LoRA), and `Ltx2.3-3DREAL-

For image-to-video or depth-related adjustments, inspect `LTXVImgToVideoConditionOnly`, `LTXVImgToVideoInplace`, `LTXAddVideoICLoRAGuide`, `DepthCrafter`, and `DownloadAndLoadDepthCrafterModel`.

Vorschlag · nicht geprüft

Verify the exact `ComfyMath` package before setup.

Vorschlag

Verify `Ltx2.3-3DREAL-strong-v2.safetensors` before setup; it is listed as a manual LoRA without a source URL or hash.

Vorschlag