Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
Simple nodes for ComfyUI
Simple nodes for ComfyUI
Package profile
Nodes in this pack
3 nodesFensTokenCounter
The FensTokenCounter node counts typed prompt tokens and also shows the active tokenizer context window tier from the connected ComfyUI CLIP object. - CLIP - Connect CLIP output (for example from a checkpoint/model loader). - Prompt Text - The text prompt to count tokens for. Supports multiline input and dynamic prompts. - Count Strategy (Advanced) - maxstream: Uses the largest tokenizer-branch count (recommended for multi-branch text encoders like SDXL/SD3). - sumstreams: Sums token counts across all tokenizer branches. - Show Token Breakdown (Advanced) - When enabled, appends a per-token breakdown (token id, decoded text, weight, and word id) for each tokenizer branch to the Details output. Off by default since it can get long for longer prompts. 1. Connect CLIP text encoder to the node. 2. Enter your prompt in the text field. 3. (Optional) Adjust advanced counting options. 4. The node outputs: - Typed token count for the prompt. - Current context-limit tier (for example 77 or 154). - Number of tokenizer chunks/windows in use. - The input prompt text (for reference). - Token Count - The number of tokens in the prompt, as counted by the connected CLIP tokenizer(s). - Context Limit - The active context window capacity tier in tokens (for example 77, 154, 231). - Chunk Count - How many tokenizer chunks/windows were needed for the prompt. - Details - A readable summary including prompt tokens, context limit, chunk count, and strategy. Includes a per-token breakdown when Show Token Breakdown is enabled. - Prompt Text - The input prompt text (multiline string). - Different tex…
Sources
1 sourceSource excerpts
1 excerptSource context: Repo Taithrah/ComfyUI_Fens_Simple_Nodes
1 input · 3 parameters
OptiEmptyLatent
The OptiEmptyLatent node calculates and generates an empty latent tensor with an optimal resolution for a given aspect ratio and megapixel (MP) target, tailored for different model types in ComfyUI. It supports both automatic (optimized) and manual (exact resolution) modes. Only preset model configurations are available. - Dimensions - Specify the aspect ratio or resolution. - Formats: - W:H (e.g. 16:9) - WxH (e.g. 1280x720) - Decimal (e.g. 1.777) - Use WxH when using exact resolution mode. - Invert - If enabled, swaps width and height (inverts aspect ratio, e.g. 16:9 → 9:16). - Optimization - TRUE: Automatically calculates the best resolution for your aspect ratio and model. - FALSE: Uses your provided resolution (must be in WxH or W:H format). - Latent Alignment - Model preset. Selects the optimization preset for the model type (e.g. SDXL, SD1.5, SD2, FLUX, Anima/Cosmos-Predict2). - Each preset has its own block size, channel count, and recommended aspect ratio range. - Batch Size - Number of latent images in the batch (higher values increase VRAM usage). 1. Optimized Mode (default): - Enter your desired aspect ratio (e.g. 16:9, 1.33, etc.). - Select the model preset (Latent Alignment). - The node will output a latent tensor with the optimal resolution for your settings. 1. Exact Resolution Mode: - Set Optimization to FALSE. - Enter your desired resolution in WxH or W:H format (e.g. 1024x768). - The node will output a latent tensor with your specified resolution, rounded to the nearest multiple of the model's block size. 1. Invert Aspect Ratio: - Enable Invert to swap wi…
0 inputs · 5 parameters
OptiEmptyLatentAdvanced
The OptiEmptyLatentAdvanced node is an advanced version of OptiEmptyLatent. It calculates and generates an empty latent tensor with an optimal resolution for a given aspect ratio and megapixel (MP) target, tailored for different model types in ComfyUI. It supports both automatic (optimized) and manual (exact resolution) modes, and allows full customization of block size, VAE scale factor, target MP, and more when using Custom mode. - Dimensions - Specify the aspect ratio or resolution. - Formats: - W:H (e.g. 16:9) - WxH (e.g. 1280x720) - Decimal (e.g. 1.777) - Use WxH when using exact resolution mode. - Invert - If enabled, swaps width and height (inverts aspect ratio, e.g. 16:9 → 9:16). - Optimization - TRUE: Automatically calculates the best resolution for your aspect ratio and model. - FALSE: Uses your provided resolution (must be in WxH or W:H format). - Latent Alignment - Model preset. Selects the optimization preset for the model type (e.g. SDXL, SD1.5, SD2, FLUX, Anima/Cosmos-Predict2, or Custom). - Each preset has its own block size, channel count, and recommended aspect ratio range. - Select Custom to manually set the following parameters: - Block Size: The minimum multiple for width/height. - VAE Scale Factor: The VAE's total downsampling factor. - Target MP: Target megapixels. - Search Range: Search range for optimization. Higher values search more possibilities (may increase calculation time). - Channels: Latent channel count (e.g. 4 for SD1/SDXL, 16 for FLUX/Cosmos-Predict2-family). - Batch Size - Number of latent images in the batch (higher values increase VR…
0 inputs · 10 parameters
