Poor multilingual & code-switching support
“The issue with Deepgram is they don’t really have solid multilingual or auto code-switching support.”
Language detection inconsistencies
“Sometimes Deepgram tagged English calls as Hindi (70% of the time).”
Accent & entity recognition issues
“Accents throw it off completely.”
Accent & entity recognition issues
“When users spell names or emails, for example firstname.lastname@gmail.com, it just can’t handle it.”
Concurrency & flexibility limits
“We quickly hit a limit when spelling names or numbers. The model just started outputting random words.”
Support & onboarding friction
“They’re not easy to work with. The process feels rigid, and they push big commitments upfront.”
Accent & entity recognition issues
“Their name and email recognition just wasn’t great.”
Reliability challenges
“If you lock the language, it won’t transcribe anything else. If you try multi-mode, it only does English and Spanish. Language detection only works on clips, not live audio. It’s frustrating.”
Support & onboarding friction
“It took months to implement. We sent thousands of emails trying to fix issues. We were just done.”
