Hello, if I were to upload audio recordings of me speaking a language not officially supported by E

[silver heart]

AI-generated Answer for cc2168

Short answer: If the language in your recordings is not one of ElevenLabs’ supported languages, Voice Changer will not reliably reproduce that language. You can upload the audio and the system may copy the voice’s timbre/tonality, but it won’t be able to actually speak the unsupported language—results are “very unpredictable,” so this isn’t recommended. (help.elevenlabs.io)

What works:

• For supported languages, any voice can speak them. For best results, use a voice cloned/trained in that same language to avoid accent drift. (help.elevenlabs.io)
• Check language support by model (e.g., Flash v2.5, Multilingual v2, Eleven v3) and pick a model that supports your target language. (help.elevenlabs.io)

Notes:

• Default/Voice Design voices are English-first and can add an English accent when used in other languages; clone a native voice to get a proper accent. (help.elevenlabs.io)
• Voice Changer converts one voice to another while preserving delivery; it doesn’t add support for languages the model can’t speak. (help.elevenlabs.io)

If you tell me the specific language, I can confirm whether it’s currently supported and which model to use. (help.elevenlabs.io)

Sources:
Can I clone my voice in a language other than English? – ElevenLabs
What languages can the AI speak? – ElevenLabs
What languages do you support? – ElevenLabs
What default voice accents do you have? – ElevenLabs
What is Voice Changer? – ElevenLabs
What languages do you support? – ElevenLabs

[cunning badge]

Would I be correct in guessing that your dataset for Irish language voice comes from the same place as Microsoft TTS?

[silver heart]

AI-generated Answer for cc2168

Short answer: I can’t confirm that. ElevenLabs has not published the source of its Irish-language training data, nor any relationship to Microsoft’s TTS datasets. Our public materials only describe training data at a high level (e.g., user-provided data with opt-out controls, plus publicly available and third‑party data), without listing language‑specific datasets. (help.elevenlabs.io)

For context, Irish (GLE) is listed as a supported language in our Eleven v3 model. (help.elevenlabs.io)

If you need an official statement, please contact our support team at [e-mail]. (help.elevenlabs.io)

Sources:
Is my data used to improve ElevenLabs’ AI models? – ElevenLabs
What languages do you support? – ElevenLabs
How can I contact support? – ElevenLabs

[cunning badge]

The problem is that all of the AI websites, and Microsoft, use data that is based on anglophone learners of Irish who do not pronounce it well. Only Abair uses native Irish speakers.

My follow on question is,

If I put the settings to a language like Russian, or maybe Lithuanian, which have phonetic inventories that are more similar to (native) Irish than English, and if I then uploaded recordings of myself speaking Irish, would those settings be able to then reproduce the sounds from my recording even though the vocabulary itself would be from a different language?