Question · Languages and accents
How do you voice type in Urdu and get English text?
Voice type in Urdu first, then translate the corrected Urdu text into English and read the result. Urdu is on Google's keyboard and Docs voice typing lists, but not on Windows voice typing. Whisper can translate Urdu speech into English in one step, but its results on a read-speech benchmark were moderate, so check every translation.
Updated October 17, 20264 min read11 cited facts, each traced to its source
Voice type in Urdu, then translate the Urdu text into English. That two-step route is the most reliable one we found, because you can read the Urdu transcript and fix it before the translator sees it. A one-step route also exists: Whisper can translate Urdu speech directly into English. Its results on the benchmark we checked were moderate, so use it for a draft and check every sentence.
For the wider method, see the guide to speech to text with translation. For Urdu dictation in general, read the question on Urdu voice typing, which covers the tools and their accuracy in more detail.
Dictate in Urdu first, then translate the text
Google added Urdu voice input in August 2017, alongside Bengali and six other Indian languages 1. You can dictate Urdu and then paste the text into a translator.
On a computer, Google's Docs help page, as archived on 10 October 2026, lists two Urdu entries, Urdu (India) and Urdu (Pakistan), among the languages and accents that voice typing works in 2. The two entries are separate, so pick the one that matches the speaker. Choosing the Urdu entry sets the language you speak. It does not, by itself, turn the output into English, so you still need the translation step.
- Choose Urdu as the dictation language in Gboard, or pick the matching Urdu entry in Google Docs.
- Dictate in short, complete sentences, with a pause at the end of each.
- Read the Urdu transcript and correct every word that looks wrong. Urdu is written right to left, so check the start and end of each line.
- Paste the corrected Urdu into a translator and compare the English with what you said.
Windows voice typing does not offer Urdu
On Windows, the built-in voice typing list does not include Urdu. Microsoft's Azure Speech table does list Urdu (India) with the code ur-IN 3, but Microsoft's own list of 46 languages and variants for Windows 11 voice typing runs from Bulgarian to Vietnamese and has no Urdu entry 4. A Windows user needs a browser tool or a third-party app for the first step. The question on Bangla, Urdu and Arabic voice typing on Windows lists the options.
Interpreter apps may not help either. Samsung's US support page lists the languages its Interpreter app supports, and Urdu is not among them 5. Check the app's language menu on your own phone before you choose it for a conversation.
Whisper translates Urdu speech into English in one step, with moderate results
OpenAI's Whisper includes Urdu in its language table 6, and its translation feature turns speech into English. OpenAI's translations endpoint outputs English only 7, so this route gives English text, not Urdu text.
On the Fleurs read-speech benchmark, Whisper large-v2 had a word error rate of 22.6 percent for Urdu 8. That is far better than its Bengali result, but it is still more than one word in five. For speech translated into English on the same benchmark, Whisper large-v2 scored 17.2 BLEU for Urdu 9. For comparison, the same model scored 22.0 BLEU for Hindi 10. Since BLEU compares the English output with reference translations, a score in this range means the English often differs from a reference. A translation can inherit the transcript's errors, which is one reason the two-step route is safer.
Read speech and conversation give different results
The benchmark numbers above are for read speech, and they can mislead you if you dictate differently. A 2024 Urdu benchmark found that Seamless-large was best on read speech, while Whisper-large was best on conversational speech 11. So the Whisper read-speech figure above is not the only result to weigh. If you talk freely, Whisper was the stronger model in that test, but check its output on your own speech before you rely on it.
It depends on the device, the speaker and the stakes
- On Android, the keyboard route is the simplest, since Urdu voice input has been available since 2017.
- On a computer, choose the Urdu (India) or Urdu (Pakistan) entry that fits the speaker, and expect the Docs route to produce Urdu, not English.
- For a recording, the one-step Whisper route can give English directly, but read the English against the audio, especially names and numbers.
- For a legal, medical or safety notice, use a fluent translator, whatever the tool.
What to do
- Use the two-step route for most work: Urdu voice typing, then translation of the corrected text.
- Speak a short Urdu paragraph and read the transcript before translating.
- Translate and compare. Read the English against the Urdu, and check names, numbers and dates separately.
- Test your own speech. Dictate two or three minutes in the accent and setting you will use, and count the errors.
- Have a fluent reviewer check anything that will be quoted, filed or acted on.
Related answers: how to voice type in Bangla and get English text, how to transcribe Tamil, Urdu or Bengali audio, and how accurate speech to text translation is. For the term, see speech translation. The dictation languages, accents and voice translation overview compares the language lists across tools.
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
Go deeper
Related questions
- How accurate is speech to text translation?
- How do you use voice input in Google Translate?
- How do you voice type in Bangla and get English text?
- Which APIs combine speech to text and translation?
- Which app translates speech to text live?
- Why is Google Translate voice not working?
- Can speech to text handle Hinglish, Spanglish and other mixed languages?
- Can Whisper translate audio?