# How Gradium Evaluates Text-to-Speech Pronunciation Accuracy

DevFeed: [How Gradium Evaluates Text-to-Speech Pronunciation Accuracy](<https://devfeed.tech/articles/the-most-accurate-multilingual-text-to-speech-by-the-numbers-81249.md>)

Original publisher: [Read original article](<https://gradium.ai/blog/word-error-rate-evaluations>)

Author: Gradium

Published: 2026-04-29T00:00:00Z

Content type: article

Language: en

Sources: [Gradium](<https://devfeed.tech/sources/gradium.md>)

Topics: [asr](<https://devfeed.tech/topics/asr.md>), [Whisper](<https://devfeed.tech/topics/whisper.md>), [Mercury](<https://devfeed.tech/topics/mercury-lang.md>), [primes](<https://devfeed.tech/topics/primes.md>)

Tags: [asr](<https://devfeed.tech/tags/asr.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [elevenlabs-flash-v2-5](<https://devfeed.tech/tags/elevenlabs-flash-v2-5.md>), [elevenlabs-multilingual-v2](<https://devfeed.tech/tags/elevenlabs-multilingual-v2.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [jiwer-alignment](<https://devfeed.tech/tags/jiwer-alignment.md>), [metric](<https://devfeed.tech/tags/metric.md>), [minimax](<https://devfeed.tech/tags/minimax.md>), [minimax-benchmark](<https://devfeed.tech/tags/minimax-benchmark.md>), [multilingual-text-to-speech-accuracy](<https://devfeed.tech/tags/multilingual-text-to-speech-accuracy.md>), [research](<https://devfeed.tech/tags/research.md>), [results](<https://devfeed.tech/tags/results.md>), [tts](<https://devfeed.tech/tags/tts.md>), [tts-benchmark-gradium](<https://devfeed.tech/tags/tts-benchmark-gradium.md>), [tts-evaluation](<https://devfeed.tech/tags/tts-evaluation.md>), [wer-multilingual-tts](<https://devfeed.tech/tags/wer-multilingual-tts.md>), [word-error-rate-tts](<https://devfeed.tech/tags/word-error-rate-tts.md>)

## AI overview

Gradium describes its word error rate (WER) evaluation for text-to-speech systems, including text normalization, transcript alignment, and results on the MiniMax Multilingual benchmark. It argues that the standard benchmark misses difficult real-world cases and that many measured errors stem from speech recognition, reference text, or normalization rather than TTS pronunciation.

## Source excerpt

How we measure WER for TTS at Gradium: text normalization, jiwer alignment, results on the MiniMax Multilingual benchmark across English, French, Spanish, Portuguese and German -- and why the standard metric is starting to saturate.