Free tool · works now
Markdown Word Counter
Accurate word count, prose character metrics, reading time, and speaking duration for Markdown notes and manuscripts.
Runs 100% in your browser. Paste Markdown articles, notes, or technical manuscripts to calculate true prose word counts, character totals, reading time, and speaking pace.
On-Device Audio Pipeline Architecture Core Architectural Principle All audio processing, circular buffer memory pools, and Whisper speech transcription execute strictly within the local macOS host boundary. Zero network sockets are initialized during recording. System Audio Loopback via ScreenCaptureKit Modern macOS releases provide low-latency audio stream taps through . By requesting an audio-only stream tap, our daemon receives raw interleaved PCM frames directly from CoreAudio without requiring third-party audio server drivers or kernel extensions. The application registers a callback that immediately copies frames into a lock-free ring buffer: Local Whisper Inference Performance Benchmarking across M2, M3, and M4 Apple Silicon architectures yields the following real-time performance characteristics: Whisper Base: 0.08x Real-Time Factor (12x faster than live speech) Whisper Small: 0.11x Real-Time Factor (9x faster than live speech) Whisper Turbo: 0.14x Real-Time Factor (7x faster than live speech) These metrics demonstrate that on-device machine learning delivers immediate responsiveness while preserving confidentiality for legal, medical, and executive discussions. For further background, consult ScreenCaptureKit Permissions and Local AI Summaries.
Why standard word counters fail on technical Markdown
Authors, researchers, and technical writers regularly need to adhere to strict word count limits when preparing manuscripts, research grant proposals, conference submissions, or executive briefing decks. However, standard browser word counters rely on simple whitespace splitting algorithms. When fed a technical Markdown document, these crude tools count every line of Swift or Python inside fenced code blocks, every key in YAML frontmatter blocks, every bracket in wikilinks, and every character in long URLs as prose words.
In a comprehensive 2,500-word engineering specification that includes configuration manifests and code examples, a naive word counter will report over 4,200 words. This leads to erroneous editorial cuts and distorted reading time estimates. Our specialized Markdown word counter strips non-prose structural syntax before executing metric calculations, giving you the true, publishable word count of your human writing.
Syntax stripping methodology and precision metrics
Our parsing engine executes a multi-pass text sanitization algorithm designed specifically for CommonMark and GitHub Flavored Markdown dialects:
- Frontmatter stripping: Detects and removes all metadata between opening and closing
---fences, discarding tags, dates, and schema properties. - Fenced code block removal: Strips triple-backtick (
```) code blocks and syntax identifiers, ensuring programming code does not inflate prose metrics. - Inline code sanitization: Removes inline backtick syntax (
`code`) while retaining surrounding sentence context. - Link anchor extraction: Extracts the human-readable text from
[anchor](url)and[[WikiLink|Alias]]links, completely discarding URLs and destination paths. - Formatting marker elimination: Strips heading hashes (
#), blockquote chevrons (>), task list checkboxes (- [ ]), and emphasis markers (**,*).
Calibrated reading time and speaking duration algorithms
In addition to word counts, pacing metrics are critical when preparing keynote talks, all-hands presentations, and engineering retrospectives. Our tool calculates silent reading duration using the internationally recognized standard of 225 words per minute. For verbal presentation planning, it applies an average conversational speaking pace of 135 words per minute, helping speakers pace their delivery accurately without rushing.
In Oyma, this level of precision is integrated into the core editing experience. Notes are rendered in real time while living as clean plain text on your local Mac disk. Learn more about our live-preview Markdown editor and discover how our bidirectional backlinks connect your research notes.
Frequently asked questions
Questions
Why do generic word counters report inaccurate counts on Markdown files?
Generic word counters treat all whitespace-separated characters as prose words. They inadvertently count fenced code blocks, YAML frontmatter keys, markdown link targets, image paths, and formatting symbols (#, *, >, ---), severely inflating word counts by 20% to 50%.
How does this tool isolate pure editorial prose from technical syntax?
Our engine parses the document structure to completely strip opening and closing YAML frontmatter, remove triple-backtick code blocks, discard inline backticks, and extract only human-visible anchor text from markdown links before computing word and character frequencies.
What reading and speaking speeds are used for duration estimates?
Reading time is calculated using an established silent reading rate of 225 words per minute. Speaking duration is calculated at 135 words per minute, matching standard executive presentation pacing and keynote delivery speeds.
Does this tool transmit my confidential drafts or notes over the internet?
No. The entire analysis runs locally in your browser tab. No text is uploaded, logged, or sent to external analytical services or third-party APIs.
How does the desktop app display document metrics?
The desktop app provides real-time word count, character count, and reading time metrics directly in the editor status bar while saving your notes as clean, unbloated Markdown files on your local filesystem.
Write distraction-free Markdown notes on your Mac
Keep your research, writing drafts, and meeting logs organized in local plain text with zero tracking in Oyma.