My personal basic HiDream Full + Flux.Dev refiner workflow

This basic HiDream Full + Flux.Dev refiner workflow is for turning text input and an empty latent image into saved images; its graph loads models, applies guidance and sampling, and decodes with a VAE.

The 19-node graph includes text encoding, latent-image setup, model loading, guidance, sampling, VAE processing, and image saving.

Named input and setup stages include CLIPTextEncode, EmptySD3LatentImage, UNETLoader, and DualCLIPLoader.

Image handling and output include VAEEncode, VAEDecode, and SaveImage.

It is listed as a Workflows entry with HiDream as its base model.

It is published by DigitalZombieAI on Civitai.

The named model files include ae.safetensors, clip_g_hidream.safetensors, clip_l_hidream.safetensors, clip_l.safetensors, flux1-dev.safetensors, and hidream_i1_full_fp16.safetensors.

The list also includes llama_3.1_8b_instruct_fp8_scaled.safetensors, t5xxl_fp16.safetensors, and t5xxl_fp8_e4m3fn_scaled.safetensors.

Named node types include CLIPTextEncode, DualCLIPLoader, EmptySD3LatentImage, FluxGuidance, KSampler, and ModelSamplingSD3.

It also includes QuadrupleCLIPLoader, SaveImage, UNETLoader, VAEDecode, VAEEncode, and VAELoader.

The named model variants include FP8 and FP16.

Before use, verify that the named model files are available and correctly named.

Suggestion · not verified

Also verify that the named node types are available in your workflow environment.

Suggestion · not verified