### Answer Yes, you can change the language mid-call using a few approaches: Option 1: Agent Swap You can use agent swap to switch to an agent configured for a different language when the user requests it. [(1) ]() *Option 2: Dynamic Variable Approach (for Conversation Flow) * If you're using Conversation Flow, you can extract the user's preferred language into a dynamic variable (e.g.,

{{user_language}}

) after the first node, then include instructions in each subsequent node like: > "Always speak in {{user_language}} throughout this node." You can also add to your global node: > "Speak only in the user's language ({{user_language}}) ." Note: With Conversation Flow agents, language switching may only apply to the current node by default—when transitioning to another node, it can revert to English unless you explicitly carry the language preference forward using dynamic variables or prompts in each node. Retell AI supports multilingual agents that can seamlessly switch between English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, and Dutch. [(2) ]()

​ To learn more, see the sources I considered: [[1] Language Switching](https://discord.com/channels/1186424376300679209/1397465772074467338) [[2] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

@AI Evy in the dashboard, if i set the language to english, the agent does not understand spanish. what i want to do is at the beginning, the agent is multilingual. not if the user response is english, then set the agent language to english. if user speaks spanish, set the language to spanish.

### Answer For your use case—starting multilingual and then locking to the detected language—here are the recommended approaches: Option 1: Use Multilingual Mode from the Start Retell AI supports Auto Language Detection that can detect and switch between 10+ languages including English and Spanish. [(1) ]() , [(2) ]() However, be aware that some users report multilingual mode can have reduced accuracy for certain languages compared to single-language mode. [(3) ]() *Option 2: Agent Transfer Approach (Recommended for Best Accuracy) * If multilingual mode doesn't provide sufficient accuracy for your needs: 1. Start with a multilingual Conversation Flow agent 2. On the first node, add two transitions: - If user speaks English → transfer to English-only agent - If user speaks Spanish → transfer to Spanish-only agent [(4) ]() This approach uses agent transfers to route to language-specific agents, which can provide better transcription accuracy than multilingual mode. , [(5) ]() Option 3: IVR/DTMF Selection If automatic detection isn't reliable enough, you can prompt the user to select their language (e.g., "Press 9 for Spanish") and use an IVR function to transfer to the appropriate language-specific agent. Key Consideration: The underlying ASR (speech recognition) is what determines whether the agent "understands" a language. When set to English-only, the ASR is optimized for English and won't reliably transcribe Spanish speech. [(6) ](

ognition-asr) Multilingual mode uses ASR that can handle multiple languages, but with potential accuracy trade-offs. [(7) ]()

​ To learn more, see the sources I considered: [[1] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) [[2] The 8 Best AI Voice Agents with Multilingual Support | Retell AI](https://www.retellai.com/blog/8-leading-multilingual-ai-voice-agents) [[3] Multilingual Agent is not working for Spanish language properly!!!](https://discord.com/channels/1186424376300679209/1430561750222372886) [[4] Discord message from @teo_bebae on 2026-02-06](https://discord.com/channels/1186424376300679209/1186462578310910112/1469268090717077608) [[5] Single Phone Number with Separate Turkish and English Agents](https://community.retellai.com/t/single-phone-number-with-separate-turkish-and-english-agents/659) [[6] Automatic Speech Recognition (ASR)](https://www.retellai.com/glossary/automatic-speech-recognition-asr) [[7] Multilingual or Bilingual Transcription](https://discord.com/channels/1186424376300679209/1230269781161279508) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer For multi-state (multi-prompt) agents, you can still implement language detection and switching. Here's how it works: Multi-Prompt Structure Multi-prompt agents use

states

where each state has its own

state_prompt

and the

general_prompt

applies across all states. [(1) ]() [(2) ]() - System prompt (with state) = general prompt + state prompt [(1) ]() Approach for Language Detection in Multi-Prompt 1. In your first state, include instructions in the

state_prompt

to detect the user's language and store it (you can use dynamic variables like

{{user_language}}

) . 2. In your

general_prompt

, add instructions like: "Always respond in the user's detected language ({{user_language}}) ." This will apply across all states. 3. Set

default_dynamic_variables

to handle cases where the language hasn't been detected yet:

json
{ "user_language": "English" }

[(1) ]() [(2) ]() Important Note on State Transitions When transitioning between states, the conversation history is retained, but state-specific prompts change. [(3) ]() The

general_prompt

persists across all states, so language instructions placed there will carry through. Alternative: Agent Transfer If multilingual mode accuracy isn't sufficient, you can use a transfer call tool to switch to a language-specific agent once the language is detected. [(4) ]()

​ To learn more, see the sources I considered: [[1] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[2] Create Retell LLM](https://docs.retellai.com/api-references/create-retell-llm) [[3] Multi-Prompt Agent State Transition - LLM Prompt Mechanics](https://discord.com/channels/1186424376300679209/1395545946066129039) [[4] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer No—if the language is set to English in the UI, the agent will not reliably understand Spanish. The underlying ASR (Automatic Speech Recognition) is what determines whether the agent "understands" a language. [(1) ]() ASR quality directly impacts every part of the AI voice agent experience—if the transcription is inaccurate, even the most advanced AI systems will misunderstand the user's intent and deliver poor results. When you select a single language (like English) , the ASR is optimized for that language and won't reliably transcribe speech in other languages. [(2) ]() Users have reported that when they select one language, transcription becomes quite weak for a second language. Your options for true multilingual support: 1. Use Multilingual/Auto-detect mode in the dashboard settings (if available for your language pair) , which uses ASR capable of handling multiple languages—though this may come with accuracy trade-offs. 2. Use Agent Transfer to route to a language-specific agent once the user's language is detected. This can provide better transcription accuracy than multilingual mode. [(3) ]()

​ To learn more, see the sources I considered: [[1] Automatic Speech Recognition (ASR)](https://www.retellai.com/glossary/automatic-speech-recognition-asr) [[2] Multilingual or Bilingual Transcription](https://discord.com/channels/1186424376300679209/1230269781161279508) [[3] Agent Transfer Node](https://docs.retellai.com/build/conversation-flow/transfer-agent-node) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Based on the available sources, there is no built-in tool call or function to dynamically change the ASR (speech recognition) language mid-call without using Agent Transfer. The language setting affects the underlying ASR, which is configured at the agent level. When you set English in the UI, the ASR is optimized for English and won't reliably transcribe Spanish speech. Your options without Agent Transfer: 1. Use Multilingual mode from the start — Set your agent to Multilingual/Auto-detect mode in the dashboard. This supports seamlessly switching between English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, and Dutch. [(1) ]() 2. Accept the limitation — If you need single-language accuracy but want to support multiple languages, Agent Transfer remains the recommended approach for routing to language-specific agents. [(2) ]() [(3) ]() Why Agent Transfer is recommended for language switching: - The destination agent has access to the full conversation history, eliminating the need for repeated customer questions. - The transition is near-instant with lower latency than other transfer methods. - All settings including language and voice will reflect the currently active agent after transfer. If Multilingual mode doesn't provide sufficient accuracy for your use case, Agent Transfer is currently the documented solution for switching language capabilities mid-call.

​ To learn more, see the sources I considered: [[1] OpenAI 4.1, Version Control, and More Languages Support](https://www.retellai.com/changelog/openai-4-1-version-control-and-more-languages-support) [[2] Agent Transfer](https://docs.retellai.com/build/single-multi-prompt/transfer-agent) [[3] Agent Transfer Node](https://docs.retellai.com/build/conversation-flow/transfer-agent-node) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.