v0.4.0

• Added on/off switch for TalkNet's autotune function. When switched on with reference audio selected, the target synthesis will copy the exact pitch of the reference audio.
• Added a language called "return_text" for support of non-English plain text languages for Tacotron 2 such as Polish, Spanish, and Portuguese.

v0.3.0

Added support for Uberduck Pipeline Tacotron 2 models. Speaker ID is stored in speaker.ntk_cfg file. Only one speaker ID per neuTalk speaker is supported so far. Model config file must be included in .ntk file under the name conf.json.

v0.2.2

Fixed issue where text with more than ten punctuation chunks would play out of order due do non-natural sorting.

v0.2.1

• Added support for English Tacotron 2 models (both character-based [eng_tt2] and ARPAbet-based [eng_tt2_arpa])
• Added definitions for French models; still does not work as I'm having trouble with the G2P
• Modified stitcher.py to include punctuation rather than replacing it all with periods
• Included SIWIS default French speaker

v0.1.1

Modified from the 0.0.1 distribution exclusive to neuTalk's Discord server.
Added tentative support for Mandarin models. Issues with Mandarin are as follows: skipping syllables; changing intonation to be incorrect. Cause of these issues is unknown.