FaceFX Documentation and support
FaceFX will give the best results when audio files are accompanied by a text file that includes a transcription of the audio file.
- Match TXT Files to Audio - Place text files in the same directory as the audio file, with the same filename but with the .txt _file extension. (Match _C:\welcome.wav with C:\welcome.txt) The TXT file will be read automatically and associated with the audio file. This is not configurable in FaceFX and it is the only way to get text associated with audio in batch analysis operations.
- Spell it Like it Sounds - When you read the text out loud, it should sound as close to the audio as possible. If words are missing in the audio, don’t include them in the text. If they are pronounced incorrectly, spell out the words like they sound in the audio.
- No Abbreviations - Do not use abbreviations in your text. Most punctuation is removed from the text prior to analysis.
- No Digits - We will try to make an educated guess at the pronunciation of any digit strings in the text, but for best results, spell out all numbers in the text as they sound
- UTF-8 Preferred - For best results, encode text files with UTF-8 Unicode. FaceFX uses ICU to determine the correct encoding and read the files.
FaceFX will filter out characters that are unsupported for a given language. See the additional language framework for more information.
- Japanese - When using Japanese text, Katakana and Hiragana must be used. Kanji are not supported. Also, it is important to have spaces in between the words, or at least where there are meaningful silences in the audio.
- Korean - When using Korean text, Hangul characters are supported and a Romanized phonetic spelling of the Hangul characters can also be used. Make sure you have the Arial Unicode MS (TrueType) font installed (ARIALUNI.TTF) in your Windows\Fonts folder so you can see the words in the phoneme bar.
- Custom - See the additional language framework for more info.