TIMIT is a well known, publicly accessible corpus that contains phonetic and lexicalic transcription of language (American English).
A sample sentence of TIMIT looks like:
She had your dark suit in greasy wash water all year (lexicalic)
h# sh ix hv eh dcl jh ih dcl d ah kcl k s ux q en gcl g r ix s ix w ao s epi w ao dx axr ao l y ih axr h# (phonetic)
So I am not talking about actual acustic data (which is also part of TIMIT), but rather about IPA that has been translated to latin alphabet. This type of representation is ideal for G2P or P2G models, which use one of the above as input and learn the mapping to the other.
Now my question: I have searched online for a long time and I've spoken to some experts as well, but nobody was aware of a German corpus that contains a phonetic as well as a lexicalic transcription. Am I missing something? Any hint is highly appreciated.