TUTORIAL : How I trained a specific french model to control my robot

[jehoshua]

My Model : (I didn’t really respect percentage LOL)

• model size (bp) : 17.9mo
• train dir : 5700 wav’s (3-5s/wav)
• dev dir : 1500
• test dir : 170
• model stopped by Early_stop param with a 0.06 Wer.

I’ve been able to use a Python tool to cut a WAV into word chunks - Longer audio files with Deep Speech

The audio outputs range from 1 second to 49 seconds. How will the longer (than 3 to 5 seconds) audio lengths affect the building of a model ?