DEV Community

AI-KSK
AI-KSK

Posted on Originally published at huggingface.co

FLUX: 2512 vs z image vs flux2 — 6 Workflows, Read From the Graphs

FLUX: 2512 vs z image vs flux2 — 6 Workflows, Read From the Graphs

This release ships 6 workflow files built around FLUX. Together they total 155 nodes. Rather than summarise the video, this write-up reads the graphs directly — what each one loads, how big its latents are, and which sampler it runs.

The workflows at a glance

Workflow Nodes Links Sampler Latent
2512 vs z image vs flux2 78 82 euler 1248x832 1f, 1328x1328 1f, 1472x1104 1f
Flux2 TurboV2加速 29 29 euler 1248x832 1f, 1472x1104 1f
Qwen Image 2512 lightx2v 14 13 — 1328x1328 1f, 1472x1104 1f
Qwen Image 2512 Wuli 14 13 — 1328x1328 1f, 1472x1104 1f
Z Image文生图 12 11 — 1472x1104 1f
zimage2步出图 8 7 — —

The node count range is wide: zimage2步出图 at 8 nodes against 2512 vs z image vs flux2 at 78. That gap is usually the difference between a single-pass preview graph and one that adds conditioning, upscale stages, or a second sampling pass.

What the graphs load

Model file Loader node
qwen_image_vae.safetensors VAELoader
qwen_2.5_vl_7b_fp8_scaled.safetensors CLIPLoader
qwen_image_2512_bf16.safetensors UNETLoader
z_image_turbo_bf16.safetensors UNETLoader
ae.safetensors VAELoader
qwen_3_4b.safetensors CLIPLoader
flux2-vae.safetensors VAELoader
flux2_dev_fp8mixed.safetensors UNETLoader
mistral_3_small_flux2_fp8.safetensors CLIPLoader

LoRAs in play

LoRA Weight
linflux_000004500.safetensors 1
Wuli-Qwen-Image-2512-Turbo-LoRA-4steps-V2.0-bf16(1).safetensors 1
Flux2TurboComfyv2.safetensors 1
Qwen-Image-Edit-2512-Lightning-4steps-V1.0-bf16.safetensors 1

Latent geometry

Distinct latent configurations across the set: 1248x832 1f, 1328x1328 1f, 1472x1104 1f.

The frame count matters more than the pixel dimensions here. A graph at 97 frames is built for roughly a four-second unit; one at 121 frames is closer to five. Sigma schedules do not transfer cleanly between them, so treat each geometry as its own tuning target.

Sampler choice

Samplers used: euler.

A plain euler is deterministic — good when the constraint is the point and noise per step would only perturb the trajectory. Ancestral variants inject noise each step and are the right pick when you want the model to explore rather than commit.

Where this set is weak

  1. Node count is not quality. The largest graph is not automatically the best one; a 40-node graph with a good prompt beats a 200-node graph with a bad one.
  2. No measured comparison in the files. The workflows carry settings, not benchmarks. Any quality claim needs your own A/B run.
  3. Model files are not included. You supply the weights; the graphs only reference filenames.
  4. Version drift. A graph pinned to a specific checkpoint revision will run on a newer one but may produce different output.

Practical order

Open the smallest graph first. It has the fewest moving parts and will confirm your models load before you spend time on the elaborate ones.

Workflows

6 graphs published on RunningHub with invite code rh-v1111:

  • 2512 vs z image vs flux2
  • Flux2 TurboV2加速
  • Qwen Image 2512 lightx2v
  • Qwen Image 2512 Wuli
  • Z Image文生图
  • zimage2步出图

Files

https://huggingface.co/AI-KSK/ksk-video-151-comfyui-workflows

Video walkthrough: https://youtu.be/CTIWLigZL1E
Bilibili: https://www.bilibili.com/video/BV1s5iqBNE6F/

Notes

  • Workflow files only. Model weights are not included and each keeps its own licence.
  • Node counts, geometries and loader filenames above are read from the published JSON files.
  • Independent AI-KSK release. Not affiliated with the model authors or RunningHub.

Optional API platform - 88API (affiliate link)

Explore 88API for access to multiple AI model APIs. Available models, routes and prices are listed on the platform; Alipay top-ups are supported.

https://88api.ai/sign-up?aff=usx3

Disclosure: I may earn a commission from eligible purchases after you register through this referral link.

Top comments (0)