Word Error Rate Calculator
Calculate the word error rate (WER) between a reference and a hypothesis transcript, the edit distance over words divided by the reference length.
Last reviewed by the Radiatus Cloud team
Calculate the word error rate between a reference and a hypothesis transcript.
Securing AI in production?
We build guardrails, governance & compliance for AI systems.
Calculate word error rate
Word error rate, or WER, is the standard metric for evaluating speech recognition and transcription systems. It is the word-level edit distance between a reference transcript and the system hypothesis, divided by the number of words in the reference. The edit distance counts the substitutions, insertions and deletions needed to turn the hypothesis into the reference, so a WER of ten percent means roughly one word in ten is wrong.
Changing one word in a three-word sentence gives one error over three reference words, a WER of about thirty-three percent.
Measuring transcription quality
WER lets you compare speech-to-text systems and track improvements objectively, and the same word-level edit distance is useful for evaluating machine translation and other text-generation tasks against references. A lower WER means a more accurate transcript. Because it can exceed one hundred percent when the hypothesis has many extra words, it is always interpreted relative to the reference length.
The word accuracy shown is simply one minus the WER, a more intuitive figure for non-specialists. All calculation happens locally in your browser.
Notes on these estimates
Because the word error rate calculator runs entirely in your browser, nothing you enter is uploaded, so you can use it with private data safely. The figures are estimates based on the values you provide and common rules of thumb, so treat them as planning guidance rather than exact measurements, and run the tool as often as you need for free.
Related tools
- AI Prompt Leakage Analyzer — Paste a system prompt and a hostile user input to see whether the prompt holds secrets and whether the input carries injection patterns. Local, instant.
- LLM Data Exposure Checker — Check if text contains data likely to be memorized or exposed by LLMs.
- AI Usage Policy Generator — Generate an acceptable use policy for AI tools in your company.
- Model Hallucination Estimator — Estimate risk of hallucinations based on task type and temperature.
Frequently Asked Questions
What is word error rate?
It is the word-level edit distance between a reference and a hypothesis transcript divided by the number of reference words, measuring transcription accuracy.
What counts as an error?
Each word substitution, insertion or deletion needed to turn the hypothesis into the reference counts as one error.
Can WER exceed 100 percent?
Yes. If the hypothesis contains many extra or wrong words, the error count can exceed the reference length, giving a WER above one hundred percent.
What is word accuracy?
It is one minus the word error rate, expressing the proportion of words effectively correct as a more intuitive percentage.
Privacy & Security
Everything runs in your browser; nothing is uploaded.
How to Use
Enter the reference and the hypothesis transcript.
Disclaimer: This tool is provided "as is" without warranty of any kind. Results are for educational and utility purposes.
Related Tools
AI Prompt Leakage Analyzer
AI SecurityPaste a system prompt and a hostile user input to see whether the prompt holds secrets and whether the input carries injection patterns. Local, instant.
LLM Data Exposure Checker
AI SecurityCheck if text contains data likely to be memorized or exposed by LLMs.
AI Usage Policy Generator
AI SecurityGenerate an acceptable use policy for AI tools in your company.