I'm getting this error when trying to create an agent: "This conversation failed with the following

[copper brook]

AI-generated Answer for snoobwa

Short answer: the validator is rejecting your agent because it’s configured as non‑English but its stored TTS model isn’t one of the v2.5 models required for non‑English Agents. Setting model_id at conversation start doesn’t help because conversation overrides don’t allow changing the TTS model—only the voice—so the validator still sees the non‑v2.5 model in the agent’s config. (help.elevenlabs.io)

What the docs say

• Agents that use languages beyond English must use Flash v2.5 or Turbo v2.5. (help.elevenlabs.io)
• By default a new agent uses Flash v2 (English‑only). When you add additional languages, the platform switches the agent to a v2.5 multilingual path. (elevenlabs.io)
• Conversation overrides support tts.voice_id, but not tts.model_id, so setting model_id at start of a conversation won’t change the stored agent model. (elevenlabs.io)
• The agent’s persisted configuration includes conversation_config.tts.model_id. You can read/update it via the Agents API. (elevenlabs.io)
• Valid model IDs include eleven_flash_v2_5 and eleven_turbo_v2_5. (help.elevenlabs.io)

How to fix

1. Check your agent’s stored model: GET /v1/convai/agents/:agent_id and inspect conversation_config.tts.model_id. If it’s not eleven_flash_v2_5 or eleven_turbo_v2_5, update it. (elevenlabs.io)
2. Update the agent: PATCH /v1/convai/agents/:agent_id with conversation_config.tts.model_id set to eleven_flash_v2_5 or eleven_turbo_v2_5. (elevenlabs.io)
3. If you’re managing languages in the UI, ensure your target language is added under Additional Languages—this automatically switches the agent to a v2.5 setup. (elevenlabs.io)
4. Don’t rely on conversation overrides to change the model; they only let you override tts.voice_id. (elevenlabs.io)

After this

[uneven beacon]

I'm trying to CREATE an agent but get this error: ""ElevenLabs agent creation error: {\n detail: {\n status: "input_invalid",\n message: '[{"type":"value_error","loc":[],"msg":"Value error, Non-english Agents must use turbo or flash v2_5.","input":{"asr":" I am setting the tts model id to turbo 2.5. This is my conversation config when creating an agent "conversation_config: {
agent: {
first_message: firstMessage,
language: userLanguage,
prompt: {
prompt: agentPrompt
},
tts: {
model_id: "eleven_turbo_v2_5"
}". I'm getting this error when posting to https://api.elevenlabs.io/v1/convai/agents/create

[copper brook]

AI-generated Answer for snoobwa

Short answer: your tts config is in the wrong place. In Agents, tts is a top‑level field inside conversation_config (a sibling of agent and asr). Because your tts was nested under agent, the server fell back to a default TTS model (e.g., turbo v2), which is English‑only, and validation then failed for a non‑English language. Move tts out of agent and set model_id to a v2.5 model.

Why this happens

• Agents accept conversation_config with top-level keys like agent, tts, asr, turn, etc. The example responses show tts at the top level, not under agent. (elevenlabs.io)
• For non‑English Agents, you must use a multilingual, low‑latency model (Flash v2.5 or Turbo v2.5). If an English‑only model is used, creation/updates can be rejected. (help.elevenlabs.io)

Fix

• Send a payload like the following (note tts is a sibling of agent):
  conversation_config: {
  agent: {
  first_message: "…",
  language: "es",
  prompt: { prompt: "…" }
  },
  tts: {
  model_id: "eleven_turbo_v2_5",
  voice_id: "your_voice_id"
  }
  }
  This structure matches the documented shape for Agents configs and overrides (agent … plus a separate tts block). (elevenlabs.io)

Tips to verify

• After creating, call GET /v1/convai/agents/:agent_id and ensure conversation_config.tts.model_id is eleven_turbo_v2_5 (or eleven_flash_v2_5). (elevenlabs.io)
• Keep agent.language as a standard code like "en"/"es" as shown in the docs’ examples. (elevenlabs.io)

Reference endpoints and model guidance

• Create agent: POST /v1/convai/agents/create. (elevenlabs.io)
• Model options and use with Agents (Flash v2.5 / Turbo v2.5). (elevenlabs.io)
• Agents language support equals Flash v2.5 and Turbo v2.5 language set. (help.elevenlabs.io)

If you

[uneven beacon]

how do i enable built_in_tools like end_call and language_detection when creating an agent through the api?