🧩 Comfy Text Stats

A lightweight ComfyUI custom node pack with text utility nodes for working directly inside your workflows:

• Character Search Replace — applies search/replace rules from a multiline mapping list using a | separator, supporting multi-character search and replace strings
• Line Batch Chunk — splits multiline text into fixed-size line batches using a batch index
• Line Context Counter — uses a visible counter that auto-increments like a seed and returns the current, previous, and next line context
• Remove Empty Lines — removes blank or whitespace-only lines from multiline text
• Text Stats — returns character, word, and line counts
• Text Load From File — loads a UTF-8 text file, skips comment lines, and returns both joined text and a line dictionary
• Text String Truncate — truncates a string to a given number of characters or words, from the beginning or end
• UTF-8 Cleaner — normalizes punctuation and removes invisible/control characters while preserving valid Unicode text

This pack is useful for text preprocessing, caption analytics, prompt inspection, and quick metadata generation directly in ComfyUI.

---

🚀 Features

✅ Simple and fast — no external dependencies
✅ Unicode-aware word detection (supports multilingual text)
✅ Line counting for multiline inputs
✅ Editable multiline character replacement rules
✅ Fixed-size multiline chunk extraction by batch number
✅ Sequential line traversal with previous/next context windows
✅ Visible line counter with seed-like auto-increment and manual reset control
✅ Empty-line cleanup for multiline strings
✅ UTF-8-safe text cleanup with smart punctuation normalization
✅ Deterministic utility nodes for stats and cleanup
✅ String truncation by character or word count, from either end ✅ UTF-8 text file loading with comment-line filtering ✅ Works with any STRING input from other nodes

---

📦 Installation

Option 1 — Via ComfyUI-Manager (recommended)

1. Open ComfyUI-Manager
2. Search for Comfy Text Stats
3. Install the node pack
4. Reload custom nodes or restart ComfyUI

Option 2 — Manual Install

1. Clone or download this repository into your ComfyUI custom node directory:

   git clone https://github.com/scofano/comfy-text-stats.git
   

2. The final structure should look like this:

   ComfyUI/
   └── custom_nodes/
       └── comfy-text-stats/
           ├── __init__.py
           ├── character_search_replace.py
           ├── line_batch_chunk.py
           ├── line_context_counter.py
           ├── remove_empty_lines.py
           ├── text_stats.py
           ├── text_load_from_file.py
           ├── text_string_truncate.py
           ├── utf8_processor.py
           ├── README.md
           └── requirements.txt
   

3. Restart ComfyUI.

---

🧠 Usage

1) Character Search Replace

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	The input text to process. Supports multiline input.
replacements	STRING	Multiline replacement rules. Each non-empty line defines one rule as `search

Outputs

Name	Type	Description
processed_text	STRING	The text after applying the replacement rules in order from top to bottom.

Default replacement list:

“|”
“|”
„|”
«|”
»|”
‘|’
‘|’
‚|’
´|’
`|’
–|-
—|, 
−|-
‐|-
‑|-
•|-
·|-
*|
 |_
…|...

Notes:

• Each non-empty line defines one rule using search|replacement format.
• Everything before the first | is the search string — it can be one or more characters (e.g. ,10|, replaces ,10 with ,).
• Everything after the | is the replacement — it can be empty (to delete the search string), one character, or multiple characters.
• Lines without a | are silently ignored.
• Leading and trailing spaces in a rule are meaningful.
• The default list is only a starting suggestion; you can edit it directly in the node.

2) Line Batch Chunk

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	Multiline text to split into chunks.
number_of_lines	INT	How many lines each output chunk should contain. Minimum: 1.
batch	INT	Which chunk to return, starting at 1. Batch 1 returns the first chunk, batch 2 the next chunk, and so on. This control auto-increments after each run like a seed.

Outputs

Name	Type	Description
chunk_text	STRING	The requested batch of lines. Returns a partial chunk if only some lines remain, or an empty string if the batch is fully out of range.

Notes:

• The node uses splitlines(), so each logical input line becomes one item in the batch calculation.
• batch = 1 with number_of_lines = 2 returns lines 1-2.
• batch = 2 with number_of_lines = 2 returns lines 3-4.
• If the final chunk has fewer than number_of_lines available, only the remaining lines are returned.
• If the requested batch starts beyond the available lines, the output is an empty string.

3) Line Context Counter

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	Multiline text to iterate through.
set_counter	INT	Visible line counter. Starts at 1, auto-increments after each run like a seed, and can be changed manually at any time.
previous_lines	INT	Number of previous lines to include. Default: 1.
next_lines	INT	Number of next lines to include. Default: 1.
current_prefix	STRING	Optional prefix/template for the current line output. Default: Current line: {line}
previous_prefix	STRING	Optional prefix/template for the previous lines output. Default: Previous {count} lines: {lines}
next_prefix	STRING	Optional prefix/template for the next lines output. Default: Next {count} lines: {lines}
all_lines_prefix	STRING	Optional prefix/template for the full text output. Default: All the content:

Outputs

Name	Type	Description
counter	INT	The visible current line counter, starting at 1.
current_line_text	STRING	The current line selected by the automatic counter, starting at line 1.
previous_lines_text	STRING	Up to the requested number of lines before the current line. Empty if none exist.
next_lines_text	STRING	Up to the requested number of lines after the current line. Empty if none exist.
combined_context_text	STRING	Previous, current, and next outputs concatenated together in that order, separated by newlines. Empty parts are skipped.
all_lines_text	STRING	The original input text with the optional all-lines prefix applied.

Notes:

• set_counter behaves like a seed-style control and defaults to the increment post-generate mode in ComfyUI.
• The counter output lets you see which line is currently selected.
• You can edit set_counter manually at any time to jump to or reset the current line.
• If set_counter is larger than the number of lines, it wraps around using the available line count.
• combined_context_text joins previous_lines_text, current_line_text, and next_lines_text using newline separators.
• If previous_lines = 0, previous_lines_text is always an empty string.
• If next_lines = 0, next_lines_text is always an empty string.
• Prefix fields are optional; if left blank, the node outputs only the line content.
• Templates can use placeholders such as {line}, {lines}, {count}, {actual_count}, {line_number}, and {total_lines}.
• If there are no previous lines or no next lines available, that output is an empty string.
• If the input text is empty, the node returns counter = 0 and empty string outputs.

4) Remove Empty Lines

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	Multiline text to clean.

Outputs

Name	Type	Description
processed_text	STRING	The input text with empty and whitespace-only lines removed.

Notes:

• Lines containing only spaces or tabs are treated as empty and removed.
• Non-empty lines keep their original content and order.
• The output is rejoined using newline characters.

5) Text Stats

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	The text input to analyze. Supports multiline input.

Outputs

Name	Type	Description
char_count	INT	Number of characters (including spaces, punctuation, and newlines).
word_count	INT	Number of words (using Unicode-aware regex).
line_count	INT	Number of lines in the input text.

6) UTF-8 Cleaner

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	The raw text input to sanitize. Supports multiline input.

Outputs

Name	Type	Description
clean_text	STRING	Cleaned UTF-8-safe text with common smart punctuation normalized and invisible/control characters removed.

7) Text Load From File

Category: Text/Utils

Inputs

Name	Type	Description
file_path	STRING	Path to the UTF-8 text file to load.
dictionary_name	STRING	Output dictionary key. Use [filename] to use the file name without extension, or enter a custom key.

Outputs

Name	Type	Description
text	STRING	Loaded file content after comment-line filtering, rejoined with newline characters.
dictionary	DICT	Dictionary containing one key mapped to the filtered line list.

Notes:

• Lines whose trimmed content starts with # are skipped.
• Remaining lines have \n and \r removed before being returned.
• If dictionary_name = "[filename]", the dictionary key is the source file name without extension.
• If the file path does not exist, the node returns an empty string and {key: []}.
• The file is read as UTF-8.

8) Text String Truncate

Category: Text/Utils

Inputs

Name	Type	Description
text	STRING	The input text to truncate. Supports multiline input.
truncate_to	INT	Target length. Positive values keep that many units from the selected side. Negative values use Python-style slicing from the selected side.
truncate_by	["characters", "words"]	Unit of truncation: by character count or word count.
truncate_from	["end", "beginning"]	Which side to slice from. "beginning" slices from the start of the text. "end" slices from the end of the text.

Outputs

Name	Type	Description
text	STRING	The truncated string.

Notes:

• When truncate_by = "characters", the string is sliced by character index.
• When truncate_by = "words", the string is split on whitespace and the requested number of words are kept.
• truncate_from = "beginning" with a positive truncate_to keeps the first N characters/words; with a negative value it keeps the last N characters/words.
• truncate_from = "end" with a positive truncate_to keeps the last N characters/words; with a negative value it removes the last N characters/words.
• If the input text is empty or the requested length is zero, the result is an empty string.

---

🧪 Example

Character Search Replace input text:

“hello”— world…*

Default rules output:

"hello",_world...

Line Context Counter input:

alpha
beta
gamma
delta

Line Batch Chunk input:

1
2
3
4
5
6
7
8
9
10

Line Batch Chunk examples:

number_of_lines = 2, batch = 1 -> 1
2

number_of_lines = 2, batch = 2 -> 3
4

number_of_lines = 6, batch = 2 -> 7
8
9
10

number_of_lines = 6, batch = 3 -> ""

Run 1 outputs (defaults, set_counter = 1, previous_lines = 1, next_lines = 1):

counter = 1
current_line_text = Current line: alpha
previous_lines_text = 
next_lines_text = Next 1 lines: beta
combined_context_text = Current line: alpha
Next 1 lines: beta
all_lines_text = All the content: alpha
beta
gamma
delta

Run 2 outputs:

counter = 2
current_line_text = Current line: beta
previous_lines_text = Previous 1 lines: alpha
next_lines_text = Next 1 lines: gamma
combined_context_text = Previous 1 lines: alpha
Current line: beta
Next 1 lines: gamma

Manual reset example (set_counter = 4):

counter = 4
current_line_text = Current line: delta
previous_lines_text = Previous 1 lines: gamma
next_lines_text = 

Zero-window examples:

previous_lines = 0  -> previous_lines_text = ""
next_lines = 0      -> next_lines_text = ""

Out-of-range counter example (set_counter = 6 on 4 lines):

counter = 2
current_line_text = Current line: beta
previous_lines_text = Previous 1 lines: alpha
next_lines_text = Next 1 lines: gamma

Text Stats input:

ComfyUI makes custom nodes easy!

Text Stats outputs:

char_count = 32
word_count = 5
line_count = 1

UTF-8 Cleaner example:

Olá… mundo – teste\u200B

becomes:

Olá... mundo - teste

Remove Empty Lines input:

alpha

   
beta

gamma

Remove Empty Lines output:

alpha
beta
gamma

Text Load From File input file (example.txt):

# comment lines are skipped
alpha
beta
# another comment
gamma

Text Load From File outputs (dictionary_name = "[filename]"):

text = alpha
beta
gamma

dictionary = {"example": ["alpha", "beta", "gamma"]}

---

🔧 Technical Details

• CharacterSearchReplace parses the replacements field line by line. Each rule uses search|replacement format: everything before the first | is the search string (any length), everything after is the replacement (may be empty). Rules are applied in order using Python's str.replace().
• Word detection uses the regex:

  \b\w+\b
  

  which matches Unicode word boundaries.
• LineBatchChunk calculates the starting line using (batch - 1) * number_of_lines and returns up to number_of_lines lines from that position.
• LineContextCounter uses a seed-like set_counter widget with control_after_generate = "increment", so the counter updates automatically after each run in ComfyUI.
• LineContextCounter selection is driven by the current input counter value and wraps around when the counter exceeds the available line count.
• RemoveEmptyLines removes lines whose content is empty after trimming whitespace.
• UTF-8 Cleaner normalizes text to Unicode NFC, replaces common smart punctuation (such as … -> ... and –/— -> -), converts common arrows like → -> ->, removes invisible/control characters, and preserves valid Unicode characters such as áéíóúç.
• TextLoadFromFile reads UTF-8 files, removes lines whose trimmed content starts with #, strips newline/carriage-return characters from each retained line, and returns both joined text and a {dictionary_name: lines} dictionary.
• TextStats counts lines using splitlines(), so an empty string returns 0 lines.
• TextStats and UTF8Processor are deterministic, and their IS_CHANGED implementations currently use input length as a lightweight cache hint.

---

💡 Use Cases

• Text preprocessing for captioning or prompt tuning
• Normalizing punctuation or replacing/removing unwanted characters
• Splitting long multiline prompts or subtitle blocks into chunked batches
• Generating text metadata before feeding into LLM pipelines
• Measuring caption sizes in dataset workflows
• Cleaning copy/pasted text before saving or reusing it in prompts
• Stepping through caption or subtitle files one line at a time with surrounding context
• Loading prompt, caption, or dictionary text files while ignoring # comment lines

---

📄 License

MIT License © 2025 — freely usable and modifiable.
Contributions welcome!