message: Acoustic data import failed: Zero transcriptions could be parsed from the given input.

Prasanna 11

In the Speech Studio, I'm trying to train a custom model.
I'm using this folder as the template for my zip file.

This is the error I get:

Number of success: 0  
Number of failure: 1  
Error message: [  
  {  
    message: Acoustic data import failed: Zero transcriptions could be parsed from the given input.  
 Error: invalid input line format: 1.wav

romungi-MSFT 46,301 Reputation points Microsoft Employee

2022-07-29T08:48:43.163+00:00

@Prasanna Are you using only the zip file(audio-and-trans.zip) from the training folder or the complete training folder to upload the file in your custom speech project?
I think the zip file works fine and the processing completes without any issues. If you edited the file and used your own data, then please ensure the format in the transcript file matches the notation specified.
Prasanna 11 Reputation points

2022-07-29T19:17:41.95+00:00
Thanks for the reply, @romungi-MSFT .
I did use my own files and posted my ffprobe output in the question. As you can see, they are clearly in the necessary format.

Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, 1 channels, s16, 256 kb/s
romungi-MSFT 46,301 Reputation points Microsoft Employee

2022-08-01T04:26:20.087+00:00
The error could be with the format in the transcript file. If the audio format is of the supported format the transcript file should also be in the specified format, i.e

<audio_file_name> <transcript_text>
Prasanna 11 Reputation points

2022-08-04T14:28:08.007+00:00

so this is the format I have:

1.wav <text>
2.wav <text>

Pretty sure it adheres to the format that I saw in the sample data.
Akihiro Inui 0 Reputation points Microsoft Employee

2024-04-15T08:33:47.4833333+00:00

Is the line tab separated? I have the same issue in Japanese.

Share via

message: Acoustic data import failed: Zero transcriptions could be parsed from the given input.

Your answer