Thank you for the complete logging output. These are the errors from your log leading to the issue as shown in the verbosePocketsphinx output:
ERROR: “file_omitted”, line 0: Line 3: Phone ‘C’ is mising in the acoustic model; word ‘CHARLAS’ ignored [....] ERROR: “file_omitted”, line 0: The word ‘CHARLAS’ is missing in the dictionary
First it can’t add the word because there is a wrong phoneme for the acoustic model and then it can’t find it for the finite state grammar because it didn’t add it to the dictionary, which is apparently an exit condition for JSGF but not for ARPA, so thanks very much for reporting this since it was also silently affecting ARPA models.
This appears to be a bug in the fallback method for Spanish phoneme generation. I could fix it locally by opening up the OpenEars file Sscribe.m and changing line 47 which currently reads as follows:
@"ch" : @" C H ",
To this:
@"ch" : @" CH ",
Can you make this change, re-compile the OpenEars framework, and confirm to me that this fixes the problem for you? If it does, can you let me know if you see the issue with any other words or phonemes so we can troubleshoot any other occurrences of this problem at the same time before I release a fix? Thank you.