#I uploaded pls file to Voice agent to adjust pronunciation in Japanese.But nothing is changed. why

1 messages · Page 1 of 1 (latest)

spark coralBOT
#
AI-generated Answer for norisuke5134

There are a few potential reasons why your .PLS file might not be working as expected to adjust pronunciation in Japanese. Here are some key points to check:

  1. Supported Models: Ensure that the model you are using supports pronunciation dictionaries. Currently, pronunciation dictionaries are supported by the "Turbo v2" and "Eleven English v1" models. If you're using a different model, the dictionary may not be applied.

  2. Phoneme vs. Alias Tags: In your example, you are using the <alias> tag to specify the pronunciation. While this is valid, ensure that the model supports alias-based replacements for the language you're working with. If not, you may need to use <phoneme> tags with the correct phonetic representation in the IPA alphabet.

  3. Case Sensitivity: Pronunciation dictionaries are case-sensitive. Double-check that the text in your project matches the case of the <grapheme> entries in your .PLS file exactly.

  4. Language and Alphabet: You specified xml:lang="ja-JP" and alphabet="ipa". Confirm that the IPA alphabet is appropriate for Japanese pronunciation in your use case. If the model expects a different phonetic alphabet (e.g., ARPAbet), the dictionary may not work as intended.

  5. File Structure: Your file structure appears correct, but ensure there are no hidden formatting issues. Validate the file against the Pronunciation Lexicon Specification to confirm compliance.

  6. Project Settings: Make sure the .PLS file has been properly uploaded and linked to your project. After uploading, the system should automatically recalculate and mark the affected parts of the project for re-conversion.

If you've checked all the above and the issue persists, you may want to test with a simpler example or verify the system logs for any error messages. If needed, you can reach out to the support team for further assistance.

crystal bane
#

i'm using Turbo v2.5.
how can i update it to v2

spark coralBOT
#
AI-generated Answer for norisuke5134

Currently, there is no manual way to switch between Turbo v2.5 and Turbo v2 models through the web interface. The model is automatically selected based on the language being used. For English, the system defaults to Turbo v2, while for other languages, it uses Turbo v2.5.

If you are working with Japanese, Turbo v2.5 will be used automatically, as it supports multilingual capabilities. Unfortunately, there is no option to force the use of Turbo v2 for non-English languages.

If you are encountering issues with pronunciation adjustments using a PLS file, it might be worth verifying if the file is correctly formatted and supported by the system. Let me know if you need further clarification or assistance!

crystal bane
#

According to below two statements, there are no ways to adjust pronunciation in Japanese?

Currently, pronunciation dictionaries are supported by the "Turbo v2" and "Eleven English v1" models. If you're using a different model, the dictionary may not be applied.

Currently, pronunciation dictionaries are supported by the "Turbo v2" and "Eleven English v1" models. If you're using a different model, the dictionary may not be applied.

spark coralBOT
#
AI-generated Answer for norisuke5134

Yes, based on the information provided, pronunciation dictionaries are supported by the "Turbo v2" and "Eleven English v1" models. If you're using a different model, such as Turbo v2.5 or any other, the pronunciation dictionary will not be applied.

For generating Japanese audio, it is recommended to use native speaker voices available in the Voice Library and present text in Katakana to improve pronunciation. However, there is no mention of direct support for pronunciation dictionaries in Japanese, so adjustments may need to rely on these alternative methods.

crystal bane
#

currently, AI reply text include Kanji(漢字) like 受付. is there any ways to force this to hiragana

spark coralBOT
#
AI-generated Answer for norisuke5134

Currently, there is no direct way to force the AI to output only in Hiragana. However, a workaround is to manually convert Kanji into Hiragana before inputting the text into the system. This can help ensure the AI processes and outputs the text in Hiragana as desired.

Additionally, the system struggles with distinguishing Kanji readings in Japanese due to automatic language detection. Using Hiragana or Katakana instead of Kanji can improve accuracy in both pronunciation and output.

crystal bane
#

how can i do this with voice AI?

to manually convert Kanji into Hiragana before inputting the text into the system