https://retellai.com/?utm_source=discord&utm_medium=seo logo
Hey STT performance is very bad is their a way to ...
# ❓・ai-evy-support
n
@AI Evy Hey STT performance is very bad is their a way to fix it?
a
Thanks for tagging me @neerajjoshi_. I'll look through relevant sources and get back to you shortly.
### Answer To improve STT (Speech-to-Text) transcription accuracy, you can implement several fixes: [(1) ]() For Special Words or Domain-Specific Terms: - Add Boosted Keywords to enhance the model's vocabulary - Support for up to 100 custom keywords - Add frequently used names, technical terms, or domain-specific words [(2) ]() For Background Noise Issues: - Adjust the denoising mode setting to see if it helps [(1) ]() - If background noise is too high, try turning on background removal [(2) ]() For Sentences Being Cut Off: - Turn on transcription mode optimized for accuracy [(1) ]() For Missing Short Responses (like "sure" or "yes") : - If background noise isn't particularly high, set the denoising mode to No Denoising - This preserves more of the raw audio signal and improves ASR accuracy for brief utterances Additional Options: - Check language settings to ensure they match the actual spoken language [(2) ]() - Switch STT provider if needed to choose a higher-accuracy speech-to-text provider The transcription quality primarily depends on the AI models used, carefully selected to balance accuracy and processing speed. [(1) ]()
​ To learn more, see the sources I considered: [[1] Increase transcription accuracy](https://docs.retellai.com/reliability/wrong-transcript) [[2] Address Metric Issues](https://docs.retellai.com/ai-qa/address-metric-issues) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
n
@AI Evy Switch STT provider if needed to choose a higher-accuracy speech-to-text provider You guys do not give option to switch stt right?
a
Thanks for the follow-up, @neerajjoshi_. I'll get back to you soon.
​ To learn more, see the sources I considered: [[1] High WER on German, Spanish](https://discord.com/channels/1186424376300679209/1221749385554169896) [[2] Create Draft Agent Version](https://docs.retellai.com/api-references/create-agent-version) [[3] Get Voice Agent](https://docs.retellai.com/api-references/get-agent) [[4] Get Agent Versions](https://docs.retellai.com/api-references/get-agent-versions) ​
### Answer You're correct. Based on community discussions, Retell AI currently does not support switching STT (Speech-to-Text) providers. [(1) ]() However, you can configure the STT mode to optimize for either latency or accuracy using the
stt_mode
parameter: [(2) ]() [(3) ]() [(4) ]() -
"fast"
- optimizes for latency -
"accurate"
- optimizes for accuracy -
"custom"
- allows custom STT configuration When set to
"custom"
, you can provide a
custom_stt_config
object. Additionally, you can use Boosted Keywords (up to 100 custom keywords) to improve transcription accuracy for domain-specific terms, names, or brands.
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.