Hi,
I’m usinga Japanese voice in elevenlabs and trying to get it to talk japanese in my bot but it just talks like it’s an english voice. any idea of what prompt I need to use to have it only pronoounce in japanese language? I have a mixture of english and japanese text throughout my knowledgebase and the subject of the bot is japanese martial arts.
Thanks
Hey Terence, anything that has to do with pronunciation or sound is likely on the ElevenLabs side. My understanding is we send text over (english language, spanish language, etc.) and the ElevenLabs voice reads it.
That being said, I am not enormously knowledgeable about this action. Maybe @stephenasuncion can comment in more depth about the ElevenLabs integration.
When I send in a text in Norwegian, it uses the v1-english model with all voice ID’s so it sounds awful. How do I make it use a different model that supports 30+ languages?
@admin_mike could you make the ElevenLabs action possible to clone, so we can make our own custom version? 
@salten_tech I’m not using Eleven Labs (I implemented my text to speech via the CoquiAi Action)… Reading the API for Eleven Labs - I would try this gem (same approach worked for me with CoquiAi):
Add this to your Trigger Prompt… ‘language_code = ’
Where is the ISO 639-1 language code. This is used to enforce a language for the Eleven Labs model. Currently only Turbo v2.5 and Flash v2.5 support language enforcement. For other models, an error will be returned if language code is provided.
The ISO 639-1 language code for Norwegian is ‘no’. This code represents the Norwegian language as a whole, encompassing its two official written standards:
Hope this helps.
Thanks for the reply 
I have tried using language_code in the trigger prompt, but it always ends up using the v1-english model_id that doesn’t support language_code.
I have also tried changing the model_id in the trigger prompt, but it does nothing.
Def. works - just tested it and confirmed:
Go here to test the action.
I entered this prompt:
Use Language_code = ‘no’, and voice_style ‘norweigen male’ and say “Use Language_code = ‘no’, and voice_style ‘norweigen male’ and say” Hei! Jeg heter Paul, og jeg er grunnleggeren av Pickaxe Studios – et digitalt verktøystudio laget spesielt for trenere innen idrett. Målet mitt er å gjøre teknologi både tilgjengelig og nyttig for trenere på alle nivåer, enten du jobber med unge utøvere, elitesatsing eller noe midt imellom.
Hos Pickaxe Studios bygger jeg digitale “Pickaxes” – små, kraftige AI-verktøy som er enkle å bruke, men gir stor verdi i hverdagen. Hver Pickaxe er skreddersydd for å hjelpe trenere med konkrete utfordringer, som treningsplanlegging, analyse av prestasjon, restitusjonsstyring eller teknikkforbedring.
I stedet for komplekse systemer eller dyre abonnementer, gir Pickaxe-verktøyene trenere raske, brukervennlige løsninger. De kan brukes direkte fra mobilen eller datamaskinen – når som helst, hvor som helst. Ingen koding kreves, ingen installasjon. Bare åpne verktøyet og kom i gang.
Mange trenere sier at det føles som å ha en ekstra assistent i teamet – én som aldri blir sliten, og som alltid har innsikt, forslag og struktur å tilby.
Jeg utvikler også spesifikke Pickaxes etter behov, slik at trenere kan få akkurat det de trenger for sin idrett, sitt nivå og sine mål.
Målet mitt er enkelt: å gi trenere mer tid til å gjøre det de er best på – trene utøvere. Teknologien skal støtte deg, ikke komplisere hverdagen din. Det er essensen i Pickaxe Studios."
Keep in mind some specific actions need particular OpenAi models to work properly. The test page uses OpenAi 3 - I prefer to use OpenaAi o3-mini for most of mine (seems reliable).
Thanks for trying, but it is the same problem.
The voice is not Norwegian, but American/English trying to speak Norwegian, so it is super difficult to understand.
@salten_tech you mean the accent not the langauge… The language is def. Norwegian!
I don’t have an Elevan Labs account but my understanding is that you need to go into your API and ensure ou have the right voice available to you. Then specify it using the tag: voice_style
If you don’t specify the correct voice it defaults to English.
Just tested it on the Eleven Labs API page and you can create your own Norwegian voice and then use the tag voice_style to use it in your prompt - worked fine for me and sounds Norwegian… not sure if EL will let me share the voice but try this link
Yeah, the English accent is barely understandable, even with subtitles 
Thanks for the link. It creates a voice, but it is… Danish 
ElevenLabs believes I want a Danish voice because Norwegian and Danish are really similar in writing, but really different when spoken 
I will play around and try to create my own Norwegian voice. Thanks for all your help 
Update: It doesn’t help using Voice Design as it only creates Danish voices. Cloning my own voice didn’t either help as the Pickaxe action uses the model “Eleven English v1”.