Incomplete transcript when the recording has long pause before conversation resume again

Question

I am having problem to have complete transcription when my recordings have trends like below:

1. There a long silent in between the conversation, i.e., there is conversation from minute 1-2 then a long pause from min2-min10 then there is conversation again from min11 - min13 and so on so forth

I am using Azure batch transcription for Thai and the transcript generated only consist of the conversation from minute 1-2 and then it will not generate the transcript for rest of the conversation

What should I do to fix it? Please advise, thanks in advance!

Answer 1

Hi Wanyee,

Thank you for contacting Microsoft Q&A Forum.

The issue you're experiencing with incomplete transcriptions due to long pauses in your recordings is a known limitation. Currently, Azure Batch Transcription does not have any parameter to configure silence time.

Steps to Fix:

Preprocess Your Audio:

Consider preprocessing your audio files to remove or reduce long pauses before submitting them for transcription. This can help ensure that the entire conversation is captured in the transcript.

Here is a sample ffmpg

ffmpeg -i input_audio.wav -af silenceremove=start_periods=1:start_duration=1:start_threshold=-30dB:stop_periods=1:stop_duration=1:stop_threshold=-30dB output_trimmed.wav

• start_periods=1: Begins trimming after one silent period.
• start_duration=1: Silence must last at least 1 second to be trimmed.
• start_threshold=-30dB: Anything quieter than -30dB is considered silence.
• stop_periods=1, stop_duration=1, stop_threshold=-30dB: Same logic applies to the end of silent segments.

Reference:

• https://learn.microsoft.com/en-us/azure/ai-services/speech-service/batch-transcription-create?pivots=rest-api#request-configuration-options

If I have answered your question, please accept this answer as a token of appreciation and don't forget to give a thumbs up for "Was it helpful"!

Best Regards,