Hi Helena,
Sorry for the delayed response.
We are currently preparing a major GSpeech update with large architectural improvements, and unfortunately some messages on our chat platform were missed during this period.
I have replied to your email with detailed instructions, configured your multilingual voice setup, and addressed aliases question.
Thanks for your patience, and we will make sure everything works correctly for your website.
Thanks! 🙂
Hi Helena,
Thank you for the positive update. We’re glad the Portuguese audio is now working well.
We’re continuing to review the additional pronunciation points shared by your team and will keep improving the audio across the website.
Thank you for your patience and cooperation.
Best regards,
Simon Poghosyan
Founder & CEO, GSpeech
Hi Helena,
Thank you for the detailed follow-up.
Based on the pronunciation cases reported by your team, we implemented language-based pronunciation aliases specifically to support cases like acronyms, technical terminology, brand names, and industry-specific words.
You can see the full workflow here:
https://gspeech.io/blog/language-based-pronunciation-aliases
We have already completed more than five review and correction rounds for the reported articles, including custom aliases, temperature handling, content regeneration, and article-specific adjustments.
At this point, further pronunciation refinement for highly specialized Portuguese technical content should be managed by your team through the language-based aliases available in the GSpeech dashboard. This is the standard workflow for ongoing content-heavy websites, because new articles can introduce new terminology and context-specific pronunciation cases over time.
You are currently using a Gemini voice model. GSpeech also provides alternative voice models from Google Cloud Text-to-Speech and OpenAI, and the pronunciation and delivery may differ between models. We recommend testing those alternatives as well to determine which voice is the best fit for your content.
We remain available to assist with the model change and to explain how to manage the aliases, but continuous article-by-article pronunciation editing for newly published specialist content will need to be handled through your own content workflow.
Best regards,
Simon Poghosyan
Founder & CEO, GSpeech
Just one additional clarification regarding voice models.
GSpeech will continue supporting new and improved AI voice models as major providers advance their text-to-speech technologies. Different models may handle the same Portuguese terminology differently.
Custom voice prompts and vibes can also help guide how certain names, acronyms, and technical terms should be read, especially as newer models provide more precise instruction control.
However, for terminology that must remain consistent across articles, language-based pronunciation aliases are still the most reliable solution. The correct spoken form can be defined once for a specific language and then applied wherever that term appears.
Thanks again for your detailed feedback.
-
This reply was modified 1 month, 1 week ago by
Simon Poghosyan. Reason: added voice vibes and prompts ability
