Problems with AI audio transcription

Hello! I’ve been importing all the MP3 files from my university Chinese textbook without any problems — it’s been a lifesaver. But recently, many of the audio files are only being partially transcribed by the automatic AI transcription: for some unknown reason, it skips the first half and only transcribes the second half of the audio. Do you have any advice on how to fix this? I really need these audios to be fully imported and transcribed in LingQ.

Here is an example of an audio it cant be fully transcribed:

1 Like

Can you try re-converting the audio to MP3 again using this online tool:
https://online-audio-converter.com/
and then give it another try?

Thanks for your reply. I’ve just tried it — it’s the same problem: the first half of the transcription is missing, while the second half is properly transcribed.

Thanks for giving it a try. Do you have issues only with those specific files downloaded from that same source? I asked out team to investigate the problem.

1 Like

I have problems with mp3 files from this Chinese book, i attach a link with all the mp3. Some of them are transcribed good and some are not.

For example the file 14.mp3 is one of the ones which cant be transcribed fully

Thanks for your time

1 Like

I got curious and tried this with Vibe . I can’t read mandarin, so let me know if the transcription below is correct for 14.mp3. I use Vibe sometimes when LingQ does not produce the result I want.

课文二
我们怎么联系大家呢?打电话?
哎,咱们写个通知贴在宿舍楼门口怎么样?
行,贴那儿大家都能看见。
不过,怎么写呢?
就写,周末天气不错,我们打算去爬山,
那里风景漂亮,空气好。
如果谁愿意参加,请和我们联系。
我们再买一些吃的带上,
就可以在那儿野餐了。
好,一定有好多同学想去。
对了,还得写上七点在学校门口集合,
我们要早一点走。
写好了,你看行吗?
你的字越来越漂亮了。
我来读一下。
通知
大家好
周末我们打算去郊外爬山野餐。
那里有山有水
空气新鲜
风景优美
有愿意参加的朋友
请和我们联系
周末天气
请,气温15度左右
集合时间,周六早上7点
集合地点,学生公寓门口
太棒了!
我忘了写联系人和电话了
没关系,加在后面就可以了

The AI transcriptions of most of my recent French audio uploads/imports have been bonkers. Much worse than the usual “hallucinations” it sometimes produces. The transcripts have always been a little quirky, but lately they are unusable.

This is my primary use for LingQ, so it’s a major pain. On these new ones, some are wrong from the beginning, and some become so after 10 minutes or so. I assume that the transcription is jumping ahead to some place much later in the audio, because after moving ahead several pages in the text it still doesn’t match the audio.

I appreciate the team’s efforts in looking into this.

1 Like

@Josu88 Thanks, we will look into this.

1 Like

thanks, thats a good transcription, i gonan google what is vibe

1 Like

Hi @sockdoll ! Could you please share an example of such a lesson with issues, or the source audio you transcribed? This will help to figure out the case.

1 Like

Unfortunately I deleted the most egregious examples and tried reuploading them. When they failed a second time I deleted them again, and deleted the audio files from my computer and moved on to other things. Fortunately I haven’t run into this in the last few days. If I do I’ll definitely contact you with the particulars.

Thanks!

2 Likes