Currently, I’m using DeepSpeech v0.4.1 with the pre-trained model provided from the link to decode TIMIT, the transcriptions are pretty bad. Is that a normal case?
