https://retellai.com/?utm_source=discord&utm_medium=seo logo
Whats the best STT you offer
# ❓・ai-evy-support
c
@AI Evy Whats the best STT you offer? Currently on Deepgram Optimized for Speed. I used Deepgram Optimized for Accuracy but I found no difference between them. Which other providers do you offer and whats the performance of each?
a
Thanks for tagging me @Ciaran. I'll look through relevant sources and get back to you shortly.
### Answer Retell AI supports multiple speech recognition (ASR) providers: Deepgram, Azure, Soniox, and AssemblyAI [(1) ]() . By default, Retell auto-routes across providers based on your agent's languages and whether it's optimized for latency or accuracy . You can also pin a specific provider manually with a custom speech-to-text config, as long as it supports every language your agent uses . The platform offers different STT modes that you can configure [(2) ]() [(3) ]() : - fast mode (optimized for speed/latency) - accurate mode (optimized for accuracy) You can set this via the
stt_mode
parameter, which accepts values like
"fast"
. For custom configurations, you can use the
custom_stt_config
object to fine-tune settings like
endpointing_ms
. Regarding performance differences between providers, the documentation doesn't provide specific benchmarks. However, each provider has different language coverage [(1) ]() : - Azure covers every supported language and acts as the broad-coverage fallback - AssemblyAI supports code-switching across its listed languages - Deepgram and Soniox support any-to-any code-switching for multilingual agents If you're experiencing transcription issues or want to explore alternatives, you might consider adjusting your
vocab_specialization
setting (options include
"general"
) or using
boosted_keywords
to improve recognition of specific terms [(2) ]() [(3) ]() .
​ To learn more, see the sources I considered: [[1] Supported languages by provider](https://docs.retellai.com/build/language-support) [[2] Create Web Call](https://docs.retellai.com/api-references/create-web-call) [[3] Get Agent Versions](https://docs.retellai.com/api-references/get-agent-versions) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
c
@AI Evy The issue is names. For example; my name is Ciaran but the most common way of spelling that is Kieran in the UK. I could have them spell it out but even then, accents cause issues and it transcribes incorrectly quite a lot
a
Thanks for the follow-up, @Ciaran. I'll get back to you soon.
### Answer For names like "Ciaran" that have common alternate spellings, here are the approaches to improve transcription accuracy: 1. Use Boosted Keywords [(1) ]() [(2) ]() Add the specific name spelling to your agent's boosted keywords list . This biases the transcription model toward recognizing that exact spelling [(1) ]() . You can add up to 100 custom keywords . However, boosted keywords work best for known names (like your team members or frequent clients) [(3) ]() . They're less effective for random names that come up unexpectedly on calls . 2. Adjust Speech Settings [(1) ]() If background noise isn't high, try setting the denoising mode to No Denoising . The denoising mode can sometimes filter out short, low-energy responses, and this preserves more of the raw audio signal, which can improve ASR accuracy for brief utterances in relatively quiet environments . 3. Switch Transcription Mode You can turn on transcription mode to be optimized for accuracy . This helps if sentences are being cut off prematurely, which impacts transcription quality . 4. Consider STT Provider and Endpointing [(4) ]() For UK accents specifically, accent variation combined with phone line compression can cause mid-sentence transcription errors . The fix typically involves: - Slightly higher endpointing settings - Better model choice - A light post-correction layer With
endpointing_ms: 1000
, the system can cut people off early, causing parts of words to get missed,
especially with UK accents . 5. Language Setting [(5) ]() [(6) ]() Ensure your language is set to
"en-GB"
for UK English , as mismatched language settings significantly increase transcription errors [(7) ]() . Keep in mind that while boosted keywords help, they won't fully work if the model isn't "hearing" the word clearly due to pronunciation or audio quality [(4) ]() .
​ To learn more, see the sources I considered: [[1] Increase transcription accuracy](https://docs.retellai.com/reliability/wrong-transcript) [[2] Step 2: Configure the basic settings](https://docs.retellai.com/build/single-multi-prompt/configure-basic-settings) [[3] transcription Errors](https://discord.com/channels/1186424376300679209/1383756141510660157) [[4] Discord message from @.boomz88 on 2026-04-16](https://discord.com/channels/1186424376300679209/1186462578310910112/1494348145805623418) [[5] Create Phone Call](https://docs.retellai.com/api-references/create-phone-call) [[6] Update Voice Agent](https://docs.retellai.com/api-references/update-agent) [[7] 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.