Transcribe Urdu audio to text, free

What makes Urdu hard to transcribe

Spoken Urdu and spoken Hindi are close to the same language, but they are written in completely different scripts running in opposite directions. The transcription decision is therefore not about sound at all — it is about which written tradition you want, which is why language selection matters more here than almost anywhere else.

What your Urdu transcript looks like

Perso-Arabic script running right to left. Short vowels are omitted as normal Urdu writing requires.

[00:00 → 00:04] سب کو سلام، آج کی میٹنگ شروع کرتے ہیں۔

[00:04 → 00:08] سب سے پہلے پچھلے مہینے کے نتائج پر بات کریں گے۔

An illustration of how Urdu transcripts are formatted — script, punctuation and timestamps. Not the output of any particular recording.

How accurate to expect

Formal and broadcast Urdu transcribes reasonably. Everyday Pakistani speech mixing Punjabi vocabulary is harder, and the training base is smaller than for Hindi.

What usually needs fixing

1. 1 Auto-detect producing Devanagari Hindi instead of Perso-Arabic Urdu for the same audio.
2. 2 Persian and Arabic loanwords with multiple accepted spellings in Urdu.
3. 3 Punjabi-influenced pronunciation in Lahore and central Punjab speech.

Getting better results

• → Never leave this on auto-detect — the model may reasonably decide the audio is Hindi and produce Devanagari.
• → Formal news Urdu is the strongest case; heavily Punjabi-influenced speech is weaker.

Who transcribes Urdu