### Answer Speech recognition issues can stem from several factors. Here are solutions to improve your agent's "hearing": Reduce Background Noise and Improve Acoustics Implement advanced noise reduction algorithms to filter out background noise and enhance speech clarity [(1) ]() . Use high-quality microphones and optimize their placement to minimize noise pickup . Improve the acoustic environment through the use of sound-absorbing materials . Optimize ASR Settings Configure the

stt_mode

parameter - set to

"fast"

for lower latency or

"accurate"

for better transcription quality [(2) ]() [(3) ]() . You can also use

custom_stt_config

for provider-specific settings . Set the

vocab_specialization

parameter to

"general"

or

"medical"

depending on your use case . Use Boosted Keywords Add domain-specific terms to the

boosted_keywords

array to bias the transcriber model toward words that are more likely to get transcribed [(2) ]() [(4) ]() [(3) ]() . This is commonly used for names, brands, and street names .

javascript
boosted_keywords: ["retell", "kroger"]

[(2) ]() Configure Denoising Mode Set

denoising_mode

to determine what denoising to use - options include

"no-denoise"

,

"noise-cancellation"

, or

"noise-and-background-speech-cancellation"

[(2) ]() [(3) ]() . Default is

"noise-cancellation"

. Language-Specific Considerations Ensure the

language

parameter matches your target language and dialect (e.g.,

"en-US"

,

"en-GB"

,

"pt-PT"

,

"it-IT"

) [(2) ]() [(4) ]() [(3) ]() . Select

"multi"

for multilingual support [(3) ]() . Recent ASR improvements have reduced Word Error Rate by 7-10 points for European languages in Accurate mode, resulting in clearer transcripts even in noisy environments [(5) ]() . Additional Troubleshooting If issues persist with short responses, try using boosted keywords for expected answers like "yes/no" [(6) ]() . For TTS provider issues, configure

fallback_voice_ids

from different providers [(2) ]() [(4) ]() [(3) ]() .

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

​ To learn more, see the sources I considered: [[1] Troubleshooting AI Voice Agent Issues | Retell AI](https://www.retellai.com/blog/troubleshooting-common-issues-in-voice-agent-development) [[2] Create Voice Agent](https://docs.retellai.com/api-references/create-agent) [[3] Get Voice Agent](https://docs.retellai.com/api-references/get-agent) [[4] List Voice Agents](https://docs.retellai.com/api-references/list-agents) [[5] Issues with Portuguese (PT-PT) Voice Agent: Latency, Audio Recognition, and Email Capture](https://community.retellai.com/t/issues-with-portuguese-pt-pt-voice-agent-latency-audio-recognition-and-email-capture/711) [[6] Italian Voice Agent Sometimes Fails to Progress Despite Clear Audio](https://discord.com/channels/1186424376300679209/1420472399270383730) ​

### Answer Since you're using the Retell dashboard, here's how to improve your agent's speech recognition directly in the UI: Access Speech Settings In your agent settings, find the Speech Settings section [(1) ]() . This is where you can configure transcription accuracy . Change Transcription Mode Set the transcription mode to "accurate" instead of "fast" [(2) ]() . This improves how well the agent understands what customers are saying . Add Boosted Keywords Use Boosted Keywords to bias the transcriber model toward specific words that are more likely to get transcribed [(3) ]() [(4) ]() . Common uses include brand names, people's names, and street names . This makes it easier for those words to get recognized . Configure Denoising Mode Adjust the Denoising Mode setting to see if it helps [(2) ]() . This can filter out background noise and improve speech clarity . Verify Language Settings Ensure your agent's language setting matches what your callers are speaking [(4) ]() . You can select the language the agent will understand (the language of user audio) . If user audio can be in multiple languages, you can select the "multi" option . Additional Settings to Review - **Responsiveness**: Lower values make the agent wait longer before responding, which can help with speech recognition in some cases [(3) ]() [(4) ](

build/conversation-flow/global-setting) - **Interruption Sensitivity**: Lower values make the agent more resilient to background speech These settings work together to improve how accurately your agent captures what callers are saying [(5) ]() .

​ To learn more, see the sources I considered: [[1] Change the STT](https://discord.com/channels/1186424376300679209/1466175091724255372) [[2] Agent issue](https://discord.com/channels/1186424376300679209/1433446566429200486) [[3] Step 2: Configure the basic settings](https://docs.retellai.com/build/single-multi-prompt/configure-basic-settings) [[4] Step 1: Configure global settings](https://docs.retellai.com/build/conversation-flow/global-setting) [[5] Discord message from @praise_121 on 2026-02-28](https://discord.com/channels/1186424376300679209/1186462578310910112/1477339685083680841) ​

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