Gigachad Nodes
A series of custom, novelty nodes for ComfyUI by Lord Winnougan
Nodes (16)
Out of VRAM mid-workflow? This passthrough node reports exactly how much you freed
Two toggles that make checkpoint loading a Sage Attention on-ramp — just install the package first
Two text encoders in one node, any dtype, GGUF included — the CLIP loader that covers the whole zoo
The KSampler that puts RES4LYF, bongmath and beta57 inside your graph — read this before you touch clown_sampler
1280×720, 97 frames, and the weird temporal compression handled for you
Diffusion model, GGUF, fp8, Sage Attention and Flux KV cache — one loader to rule the memory trade-offs
Stack a dozen LoRAs with per-row toggles, search, and a whole lot of gold sparkles
The negative-prompt toggle that fixes Flux2, Kl-F8-Anime2 and Z-Image Turbo — and why it matters
Point it at a .txt file and it steps through your prompts automatically — the batch-workflow workhorse
WIDTH, HEIGHT and a ready-made empty latent
RTX Video Super Resolution as a node — real-time pixels, but read this before you expect 'more detail'
Noise scaling, CFG rescale, and live previews without extra nodes
Display any string right on the node — the debugging tool that doubles as a workflow label
VAE decode with tiled fallback and video-frame flattening — the last node in every workflow
Turn a starting image into a latent — with tiled encoding for the big ones and video support
A VAE loader with a 'pixel_space' option you'll almost certainly be confused by (until now)
⚡ Gigachad Nodes — ComfyUI Custom Node Suite
<img width="1200" height="896" alt="Woman_holding_golden_sign__GIGAC…_202605211143" src="https://github.com/user-attachments/assets/54bb8be0-1de2-4f2a-972c-21e212976916" />A collection of custom, novelty nodes for ComfyUI by Lord Winnougan
🎨 AI Art · Video · LLM Workflows · Generative Pipelines
📦 Included Nodes
| Node | Description | |---|---| | Gigachad Model Loader | Load diffusion models with Gigachad energy | | Gigachad Checkpoint Loader | Checkpoint loading, streamlined | | Gigachad CLIP Loader | Load CLIP models cleanly | | Gigachad Prompt Encoder | Encode prompts like a chad | | Gigachad Prompt File Reader | Read prompts from external text files | | Gigachad KSampler | Feature-packed sampler with ClownSampler/RES4LYF support and bongmath | | Gigachad Sampler Custom Advanced | Advanced custom sampling controls | | Gigachad Power LoRA Loader | Load multiple LoRAs with power | | Gigachad Resolution Picker | Pick standard resolutions fast, outputs WIDTH + HEIGHT + LATENT | | Gigachad LTX Resolution Picker | Resolution picker tuned for LTX Video | | Gigachad VAE Loader | Clean VAE loading node | | Gigachad VAE Encode / Decode | VAE encode and decode utilities | | Gigaresolution (RTX Super Res) | RTX-powered super resolution upscaling | | Gigachad Cache Cleanup | Free up VRAM/RAM with one node, reports timing and free VRAM | | Gigachad Show Text | Display text output inline in the graph |
🚀 Installation
Option 1 — ComfyUI Manager (Recommended)
Search for comfyui-gigachad in the ComfyUI Manager and install directly.
Option 2 — Manual Install
-
Navigate to your ComfyUI custom nodes folder:
cd ComfyUI/custom_nodes -
Clone this repo:
git clone https://github.com/Winnougan/comfyui-gigachad.git -
Restart ComfyUI.
No extra dependencies required — pure ComfyUI native.
🖥️ Requirements
- ComfyUI
- Python 3.10+
- A GPU that doesn't fear greatness
📖 How to Use the Nodes
All Gigachad nodes live under the Gigachad category in the node menu. Right-click the canvas → Add Node → Gigachad.
🔧 Basic txt2img Workflow
A typical Gigachad pipeline looks like this:
Gigachad Checkpoint Loader
↓
Gigachad CLIP Loader → Gigachad Prompt Encoder (positive + negative)
↓
Gigachad Resolution Picker → Gigachad KSampler
↓
Gigachad VAE Decode → Save Image
🖼️ Gigachad Resolution Picker
Outputs WIDTH, HEIGHT, and an empty LATENT — wire all three directly into the KSampler.
| Input | Description |
|---|---|
| width | Image width in pixels (default 1024, step 8) |
| height | Image height in pixels (default 1024, step 8) |
| batch_size | Number of images to generate at once |
Tip: Use the LTX Resolution Picker instead if you're running LTX Video workflows — it uses LTX-specific aspect ratios.
⚡ Gigachad KSampler
The star of the show. Drop-in replacement for the standard KSampler with extra firepower.
| Input | Description |
|---|---|
| model | Connect your loaded model |
| positive / negative | Conditioning from Prompt Encoder |
| latent_image | Connect from Resolution Picker or VAE Encode |
| seed | Generation seed |
| steps | Number of sampling steps (default 20) |
| cfg | Classifier-free guidance scale (default 7.0) |
| sampler | Standard ComfyUI sampler list |
| clown_sampler | RES4LYF/ClownSampler list — overrides sampler when not set to none |
| scheduler | Noise schedule (includes beta57, linear_quadratic extras) |
| denoise | Denoising strength (1.0 = full generation, lower = img2img) |
| bongmath | Toggle high-precision denoising for Flux/DiT models |
| bongmath_cfg_scale | CFG scale inside bongmath (independent of outer cfg) |
| bongmath_scale | Bongmath noise scale / step multiplier |
| sigmas (optional) | Override the scheduler with a custom sigma schedule |
| options (optional) | RES4LYF OPTIONS block for advanced ClownSampler control |
Outputs: latent and denoised_output — use latent for normal workflows, denoised_output for chaining into further processing.
Tip: If you're not using RES4LYF, leave
clown_sampleronnoneand it'll use the standard sampler list as normal. Bongmath is specifically for Flux/DiT models — leave it off for SD1.5/SDXL.
🧹 Gigachad Cache Cleanup
A passthrough node you can drop anywhere in your workflow to free VRAM between heavy operations. Accepts and returns any type — it won't break your connections.
| Input | Description |
|---|---|
| any_input | Connect anything (optional) — it passes straight through |
| empty_cache | Toggle torch.cuda.empty_cache() |
| gc_collect | Toggle Python garbage collection |
After running, it displays elapsed time in milliseconds and your current free/total VRAM in the node UI.
Tip: Place one after your KSampler and before upscaling to reclaim VRAM before the next big operation.
📄 Gigachad Prompt File Reader
Load prompts from a .txt file on disk instead of typing into the graph. Great for batch workflows and prompt libraries.
Connect its output to the Gigachad Prompt Encoder positive or negative input.
📝 Gigachad Show Text
Displays any text string directly in the graph as a read-only node. Useful for debugging prompt outputs, displaying metadata, or just labeling sections of a complex workflow.
🔍 Gigaresolution (RTX Super Res)
Upscale your generated images using RTX Super Resolution. Connect the output image from your VAE Decode into this node for a clean upscale pass.
Note: Requires an NVIDIA RTX GPU with NIS/RTX Super Resolution support.
💡 Example: Minimal Workflow
Gigachad Checkpoint Loader
↓ MODEL, CLIP, VAE
Gigachad CLIP Loader (if needed)
Gigachad Prompt Encoder
← CLIP
← positive prompt text
← negative prompt text
↓ CONDITIONING (positive + negative)
Gigachad Resolution Picker
↓ WIDTH, HEIGHT, LATENT
Gigachad KSampler
← MODEL
← positive CONDITIONING
← negative CONDITIONING
← LATENT (from Resolution Picker)
↓ latent
Gigachad Cache Cleanup ← (optional, drop here to free VRAM)
↓ passthrough latent
Gigachad VAE Decode
← VAE
← latent
↓ IMAGE
Save Image
❤️ Support
If these nodes save you time or spark something cool, consider supporting on Patreon — exclusive workflows, nodes, and LLM setups drop there first.
📄 License
Apache 2.0 — use it, build on it, don't be lame about it.