Deepgram

Deepgram is a developer-focused voice AI platform providing low-latency speech-to-text, text-to-speech, and a Voice Agent API. Its Nova-3 model transcribes pre-recorded and streaming audio in 40+ languages with speaker diarization, smart formatting, automatic language detection, and keyterm prompting. Pricing is usage-based and among the most affordable in the category, with $200 in free starting credits and volume discounts via prepaid annual plans. Deepgram raised a $130M Series C at a $1.3B valuation in January 2026 and is best suited for teams building voice-enabled products at scale.

Deepgram builds voice AI infrastructure for developers and enterprises, centered on high-accuracy, low-cost speech-to-text. Its Nova model family (Nova-3 is the current flagship) transcribes pre-recorded and streaming audio in more than 40 languages, with features like speaker diarization, smart formatting, automatic language detection, and keyterm prompting to boost recognition of domain-specific vocabulary. Beyond transcription, Deepgram has expanded into text-to-speech and a Voice Agent API that stitches speech recognition, language models, and speech synthesis into real-time conversational applications. The platform is priced on usage, with per-minute rates that are among the lowest in the category, plus free credits to start and volume discounts through prepaid annual commitments. In January 2026 Deepgram raised a $130 million Series C at a $1.3 billion valuation, underscoring its position as a well-funded, developer-first alternative to hyperscaler speech services. It is a strong fit for teams building voice-enabled products at scale who want transcription accuracy, speed, and predictable cost.