AI-generated Answer for naquedam
Short answer: always send ISO 639-1 two-letter language codes (e.g., fr, ar, hi). The three-letter codes you see in the v3 language list (fra, ara, hin, …) are not what the API expects.
-
Old Text-to-Speech endpoint (POST /v1/text-to-speech/:voice_id)
Use the language_code field with ISO 639-1:
{
"text": "...",
"model_id": "eleven_multilingual_v2" | "eleven_v3",
"language_code": "fr" // or "ar", "hi", etc.
}
The API explicitly documents language_code as “ISO 639-1.” (elevenlabs.io)
-
Text-to-Dialogue (POST /v1/text-to-dialogue)
Same parameter and format:
{
"inputs": [{ "voice_id": "...", "text": "..." }],
"language_code": "fr" // or "ar", "hi", etc.
}
The API also specifies ISO 639-1 here. (elevenlabs.io)
Notes:
- The v3 docs list supported languages with 3-letter tags (e.g., fra, ara), but that list does not change the API requirement; the request parameter is still ISO 639-1. (elevenlabs.io)
- There is no separate “country code” parameter; accent comes from the voice you choose. If needed, pick a voice trained in your target accent or clone one; language selection via API remains language_code (ISO 639-1). (help.elevenlabs.io)
Sources:
Create speech | ElevenLabs Documentation
Create dialogue | ElevenLabs Documentation
Models | ElevenLabs Documentation
How do I select the language and accent? – ElevenLabs
nova zephyr
fast holly