Ramon Guthrie
83de1cc040
RBG Smart Seed Variance — Minor fixes 2.2.1
Minor text updates and small fixes
Previous major updates
- Add Krea2 (SingleStream) model type with band-aware noise handling across
the 12-layer MMDiT tap stack, including a dedicated base-embedding
rebalance pass to boost expression, attention detail and prompt following.
- Add ERNIE-Image to model-specific adjustment presets, alongside
refined multipliers for Flux, Chroma HD, SDXL and Wan2.2.
- Add interactive token protection inspector (JS): click-to-toggle individual
tokens directly on a live visualisation of the protection mask, replacing
manual range typing as the primary workflow.
- Add "Custom Regions" and "Random Regions" protect modes — custom accepts
explicit token/range syntax (e.g. "0-5,15-20"), random seeds a procedural
protection mask.
- Add live protection mask feedback from backend to frontend via onExecuted,
so the inspector reflects the actual tokens used in the last run rather
than a static preview.
- Add Variance Schedule (Composition Lock 🔒) — control how noise intensity
changes across the sampling timeline with 5 modes: constant, decreasing
(fade out), step_cutoff (hard switch at a step), hard_lock (zero variance
until cutoff), and tiered_release (multi-phase staged unlock)
- Add cutoff_step, total_steps, and cutoff_strength controls — map a step
number to a percentage of the sampling timeline and set the noise
multiplier applied after the cutoff point, so composition can be locked
early while detail variance continues later.
- Expand direction shift library from the original artistic set to 26+
patterns, including Face-Variance Expansion, Visceral Expression & Grit
(Krea2-tuned), Semantic Drift, Structural Lock, Cinematic Framing, Texture
Lift, Landscape Depth, Group Diversity, Expression Variance, Motion Blur,
Microscopic, Cosmic, and Bone Anatomical Coherence.
- Add dynamic widget visibility — protect_regions field now shows/hides
automatically based on the selected protect_mode.
- Fix word-index vs BPE-index mismatch in token protection by adding
bidirectional BPE-to-word mapping, so protected ranges align correctly
with actual tokenizer output instead of word-level approximations.
ComfyUI-RBG-SmartSeedVariance 🌱
Advanced Seed Diversity Enhancement for ComfyUI ✨
Some generative models (like Z-Image Turbo and Qwen-Image) suffer from limited seed variance — changing the seed produces only subtle variations or nearly identical outputs. The RBG Smart Seed Variance node solves this by intelligently injecting controlled noise into text embeddings during generation, creating meaningful diversity while preserving your prompt intention.
Feature List 🚀
-
7 Intelligent Presets: Pre-configured variance levels from Subtle to Wild:
- 🌱 Subtle - Gentle diversity for fine-tuning
- 🌿 Balanced - Sweet spot for most use cases
- 🪴 Creative - Unlock more artistic variations
- 🌳 Bold - Significant structural changes
- 🌴 Wild - Maximum diversity for exploration (Note this might break your prompt, use with caution!)
- ⚙️ Custom - Fine-tune with percentage slider (0-100%)
-
Model-Specific Optimization: Automatic adjustment for your architecture:
- Krea2, Z-Image Turbo, Ernie-Image, Qwen-Image, Flux (Dev/Schnell), Wan, Chroma HD, SDXL and more!
-
25 Direction Shift Patterns: Apply structured artistic biases instead of pure random noise.
-
7 Spatial Fade Curves: Control how noise fades across the embedding space.
-
Flexible Noise Injection Timing: Control when variance is applied.
-
Prompt Token Protection: Preserve specific parts of your prompt from noise.
📥 Installation
- Clone this repository into your
ComfyUI/custom_nodesdirectory:git clone https://github.com/RamonGuthrie/ComfyUI-RBG-SmartSeedVariance.git - Install the required dependencies:
pip install -r requirements.txt - Restart ComfyUI.
The node will appear under the RBG Suite/Advanced category.
💡 Pro Tips
- Start conservative: Begin with "Balanced" preset and adjust up/down based on results
- Experiment with directions: Each artistic direction creates unique aesthetic shifts - try them all!
- Protect your keywords: Use "First Half" protection if important prompt concepts are being overridden
- Track your seed: Use consistent seeds to compare variance effects side-by-side
- Model matters: Always select the correct model type for optimal results and to avoid over-correction
- Combine strategies: Mix direction shifts with fade curves and noise injection timing for sophisticated effects
- Use custom mode: Fine-tune with the slider when presets don't match your needs exactly
- Chain more than one: For advanced effects, you can chain multiple Smart Seed Variance nodes.
- Share settings with the Community: If you find a killer setting combination Export them and share with the open-source community.
Watch the Demo video 📺
https://github.com/user-attachments/assets/e33b7139-8979-44f1-ae70-dd46d9b1a91e
🐛 Troubleshooting
Output looks exactly the same?
- Increase preset to "Bold" or "Creative"
- Check that the node is connected to your conditioning
- Verify model type is correct for your actual model
- Try a different seed value
Quality degraded or image broken?
- Reduce preset to "Subtle"
- Enable prompt protection ("First Half" or "First Quarter")
- Switch direction shift to "🚫 None" to use pure random
- Try "Ending Steps" to limit variance timing to fine details only
Getting strange/unexpected outputs?
- Reduce shift_strength to 50-70%
- Try a different direction shift pattern
- Verify ComfyUI version compatibility
Usage 🚀
To use the RBG Smart Seed Variance node, connect it between your KSampler and the Conditioning input. This allows the node to modify the conditioning based on your chosen variance settings.
Contributing ❤️
Contributions are always welcome! If you have any suggestions, improvements, or new ideas, please feel free to submit a pull request or open an issue.
License 📜
This project is licensed under the GNU General Public License v3.0 (GPL-3.0) — see the LICENSE file for the full text.
What this means in practice:
- ✅ You can use, study, modify, and redistribute this code — including for commercial ComfyUI workflows.
- ✅ You can fork it and build on it.
- ⚠️ If you distribute a modified version, you must release your changes under GPL-3.0 too, and make the source available to whoever you distribute it to.
- ⚠️ You must keep the copyright notice and this license intact, and clearly mark what you changed.
- ❌ You cannot fold this code into a closed-source or proprietary node pack. This matches the license of ComfyUI itself, so it stays fully compatible with the ecosystem this suite is built for.
Prior to [12/11/2025], this repository was released under the MIT License. Code distributed under the old MIT license terms before that date remains available under MIT for anyone who already forked/cloned it then; all new releases and changes are GPL-3.0.
This project is licensed under the MIT License.