ComfyUI_Fens_Simple_Nodes Introduction
ComfyUI_Fens_Simple_Nodes is a collection of custom ComfyUI nodes that helps you manage three common parts of an image-generation workflow: prompt length, image dimensions, and LoRA compatibility for certain Anima models.
The nodes serve different purposes, so you can use only the ones your workflow needs:
- Token Counter shows how many tokens your typed prompt uses and how full the connected text encoder’s context window is.
- Optimal Empty Latent creates an empty latent at a resolution suited to your chosen aspect ratio and model preset. An advanced version gives you more control over how that resolution is chosen.
- Anima LoRA Remap prepares an Anima 2.9B or 3.8B model to use LoRAs trained on the smaller Anima 2B base model more effectively.
How ComfyUI_Fens_Simple_Nodes Works
In ComfyUI, a node takes information from your workflow, performs a specific task, and passes its result to the next node. These nodes fit into that same pattern.
Token Counter reads your typed prompt through the connected text encoder. Think of the encoder as the part of the workflow that turns words into units called tokens. The counter helps you see how much of the encoder’s available context your prompt occupies. It also handles prompts split with BREAK and workflows with more than one encoder.
Optimal Empty Latent creates the starting canvas that an image-generation model fills in. You supply a shape, such as 16:9, and the node chooses suitable width and height values using its preset or the settings you provide. You can also connect a model so the node takes latent channel and VAE-scale information directly from it.
Anima LoRA Remap addresses a mismatch between model sizes. A LoRA trained for Anima 2B may load on Anima 2.9B or 3.8B without warning, even though it does not line up with the intended layers. The remap node passes the larger model onward in a form that lets a downstream LoRA loader apply those Anima 2B LoRAs correctly.
ComfyUI_Fens_Simple_Nodes Features
Token Counter
Use Token Counter when you want to check prompt length while working with your connected text encoder. It counts typed prompt tokens and indicates how full the active context window is.
The node supports fixed-window and unbounded encoders, including CLIP, T5XXL, and Qwen3-family encoders. It also handles multi-encoder models and BREAK splitting. This makes it useful when you are experimenting with a long prompt or comparing how a prompt behaves in different model workflows.
Token Counter example (https://raw.githubusercontent.com/Taithrah/ComfyUI_Fens_Simple_Nodes/refs/heads/main/examples/TokenCount.webp)
The original documentation also provides a WebUI-Forge token-count comparison image (https://raw.githubusercontent.com/Taithrah/ComfyUI_Fens_Simple_Nodes/refs/heads/main/examples/ForgeCount.webp).
Optimal Empty Latent
Use Optimal Empty Latent to create an empty latent with dimensions suited to your intended image shape. You can enter an aspect ratio in whichever of these forms is easiest to work with:
- A ratio, such as
16:9 - Dimensions, such as
1920x1080 - A decimal, such as
1.7778
You can let the node choose a resolution using a model preset or provide your own target. You can switch between portrait and landscape and create a batch when you want to generate multiple images. Connecting a MODEL is optional; when one is connected, the node takes its latent channel count and VAE scale from that model.
For example, 16:9 describes a wide image, while switching to portrait gives you a taller canvas. Changing the resolution affects the size of the latent passed into the rest of your workflow.
Optimal Empty Latent example (https://raw.githubusercontent.com/Taithrah/ComfyUI_Fens_Simple_Nodes/refs/heads/main/examples/OptimalEmptyLatent.webp)
Optimal Empty Latent (Advanced)
The advanced node includes the same core options and adds a Custom preset. It lets you set the block size, VAE scale factor, target megapixels, search range, and channel count yourself.
This version is useful when an existing preset does not match the workflow you are building. If you do not need to control those values, the standard Optimal Empty Latent node is the simpler choice.
Anima LoRA Remap
Anima 2.9B and 3.8B are expanded versions of the Anima 2B base model, anima-base-v1.0. A LoRA trained on Anima 2B can appear to load on a larger version but produce a weak, unexpected, or absent effect because its layers do not align correctly.
There are two remap nodes: one for using Anima 2B-trained LoRAs with Anima 2.9B, and one for using Anima 2B-trained LoRAs with Anima 3.8B. Choose the node that matches the larger model in your workflow. Place it immediately after the model loader and before any LoRA loader or LoRA stack node. You can keep using your existing LoRA loader or stacker downstream.
ComfyUI_Fens_Simple_Nodes Models
The extension provides model presets for Optimal Empty Latent, not downloadable image-generation models. Presets are available for SD1, SD2, SDXL, SD3, FLUX.1, FLUX.2, Anima/Cosmos-Predict2, Qwen-Image, Qwen-Image 2.1, HunyuanImage 2.1, MOSS-Image, and Stable Cascade (C-prior).
Choose the preset that corresponds to the model workflow you are using so the node can select an appropriate starting resolution. If you need to specify the underlying values yourself, use Optimal Empty Latent (Advanced) and its Custom preset.
The Anima LoRA Remap nodes have a separate role: they support LoRAs trained on Anima 2B when you use them with Anima 2.9B or Anima 3.8B. They do not provide those models or LoRAs.
Troubleshooting ComfyUI_Fens_Simple_Nodes
An Anima 2B LoRA loads, but its effect is weak or missing
- Check whether your workflow uses Anima 2.9B or 3.8B with a LoRA trained on Anima 2B.
- Add the remap node that matches your larger Anima model.
- Connect it directly after the model loader.
- Connect your LoRA loader or stacker after the remap node, then try the workflow again.
The empty latent does not have the shape you expected
- Check the aspect ratio you entered. For a wide image, for example, use
16:9,1920x1080, or1.7778. - Check whether the portrait or landscape option matches your intention.
- Check that the selected preset matches your model workflow, or set a custom target using the advanced node.
Frequently Asked Questions
Do I need to connect a model to Optimal Empty Latent?
No. A MODEL connection is optional. When connected, the node uses latent channel and VAE-scale information from that model.
Do I need a different LoRA loader to use Anima LoRA Remap?
No. The remap node is designed to work before the LoRA loader or stacker you already use.
Does the Anima remap apply to every LoRA?
The documented remap nodes are specifically for LoRAs trained on Anima 2B when used with Anima 2.9B or 3.8B.
Learn More about ComfyUI_Fens_Simple_Nodes
- Project documentation on DeepWiki (https://deepwiki.com/Taithrah/ComfyUI_Fens_Simple_Nodes) offers another way to explore the project.
- Project issues let you review reported problems and share a reproducible issue.
- The illustrated Token Counter example (https://raw.githubusercontent.com/Taithrah/ComfyUI_Fens_Simple_Nodes/refs/heads/main/examples/TokenCount.webp) and Optimal Empty Latent example (https://raw.githubusercontent.com/Taithrah/ComfyUI_Fens_Simple_Nodes/refs/heads/main/examples/OptimalEmptyLatent.webp) show the nodes in use.
