facok

facok / comfyui-meancache-z

Public

Unofficial MeanCache for Z-Image (ComfyUI)

31
3
100% credibility
Found Feb 05, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A ComfyUI custom node that speeds up Z-Image AI image generation by intelligently skipping stable steps using average velocity predictions, based on MeanCache research.

How It Works

1
📰 Discover faster AI images

You hear about a simple add-on that makes your AI image generator create pictures much quicker without losing quality.

2
📥 Get the speed booster

Download the tool and drop its folder into your image app's special add-ons spot.

3
🚀 Open your image maker

Fire up your favorite AI image program where you usually create artwork.

4
🔗 Hook up the booster

Drag your image generator into the special speed booster box to supercharge it.

5
Pick your speed style
⭐
Quick preset

Select Quality, Balanced, Speed, or Turbo for instant setup.

🔧
Personal tweak

Adjust dials for skip amount, start point, and safety levels to match your needs.

6
▶️ Hit generate

Connect everything and press go to start making images.

🎉 Faster masterpieces

Watch your beautiful images appear 1.5 to 2 times quicker, with a fun summary of the speedup in the log.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 31 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is comfyui-meancache-z?

This Python ComfyUI custom node delivers training-free speedup for Z-Image Flow Matching models by intelligently skipping stable sampling steps using cached average velocities. Drop it into your custom_nodes folder, wire your Z-Image model through the MeanCache node, pick a preset like Balanced or Turbo, and get a patched model with 1.4x-2.0x faster inference while holding image quality. It spits out a console summary of steps skipped and speedup achieved, making it dead simple for ComfyUI workflows.

Why is it gaining traction?

Unlike generic samplers or heavy fine-tunes, it's a zero-training plug-in tailored for Z-Image that maintains ODE trajectory accuracy via presets—no tweaking required unless you want custom thresholds. Developers dig the PSSP scheduling and adaptive skipping that hit real speedups without artifacts, plus CPU/CUDA cache options to fit VRAM limits. As an unofficial take on UnicomAI's MeanCache, it slots right into ComfyUI like other github unofficial tools, delivering measurable gains in seconds.

Who should use this?

ComfyUI power users generating with Z-Image models who hate waiting 30+ steps per image. AI artists batching high-res outputs on consumer GPUs, or node graph tinkerers optimizing Flux-like flows. Skip if you're not on Z-Image or prefer manual sampler tweaks.

Verdict

Grab it if Z-Image slowdowns bug you—solid docs and MIT license make testing low-risk, with presets nailing 1.7x balanced speedups out the gate. At 21 stars and 1.0% credibility, it's early beta (no tests visible), so watch for edge cases in CFG or LoRAs, but promising for comfyui meancache fans.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.