4 or 9 steps, Sampler chains & noise types chains - Flux, Mochi, PDXL, MoGe, LTXV, Others

This workflow provides 4- or 9-step sampler and noise-type chains for Flux, Mochi, PDXL, MoGe, and LTXV, with text-conditioning and image-to-video paths and save nodes for images, animated WEBP, AVIF, and latents.

The 82 listed nodes cover model loading, text conditioning, sampler and scheduler selection, image-to-video processing, decoding, and saving.

Text-conditioning paths are represented by CLIPTextEncode and LTXVConditioning, while LTXVImgToVideo is listed for image-to-video processing.

Image Saver, SaveAnimatedWEBP, SaveAvif, and SaveLatent are the listed output nodes.

The workflow is published by Ektlinux and labeled as type Workflows with base model Other.

Before using it, check that the named model files and node components are available in your environment.

Suggestion · not verified

Named model files are clip/t5xxl_fp8_e4m3fn.safetensors, clip/vpony-v01-sdxl-clipG.safetensors, clip/vpony-v01-sdxl-clipL.safetensors, hsxl_temporal_layers.f16.safetensors, unet/ltx-video-2b-v0.9-unet.safetensors, unet/vpony-v01-sdxl-unet.safetensors, and vae/ltx-video-2b-v0.9-​​

Named node components include ADE_AnimateDiffLoaderGen1, ADE_StandardStaticContextOptions, BetaSamplingScheduler, BlehInsaneChainSampler, BSZColoredLatentImageXL, CLIPLoader, CLIPNegPip, CLIPTextEncode, ClownsharKSampler, ConditioningZeroOut, DualCLIPLoader, FreeUExtreme, FreeUE极

Additional named node components include MultiplePathsInput, MUSimpleWildcard, NoisyLatentLike, OverrideCLIPDevice, Power Lora Loader (rgthree), Qwen2_VQA, RungeKuttaSampler, SamplerDPMAdaptative, SamplerLCMCustom, SamplerSonarDPMPPSDE, SaveAnimatedWEBP, SaveAvif, SaveLatent,Sh​​

The title identifies 4-step and 9-step options, sampler chains, and noise-type chains.

Listed sampler and noise controls include KSamplerSelect, BetaSamplingScheduler, LaplaceScheduler, and SplitSigmas.

It is described as work in progress, and models marked 📦 are described as archived and not yet updated.

Verify how to embed nodes into images or videos and whether OmniGen/CogVideoX/LLM/t2a/AnimateDiff(gpu) can load through CustomSamplerAdvanced before relying on those capabilities.

Suggestion · not verified