Deepgram

Deepgram is a voice AI company that provides speech APIs built for developers building production applications rather than end users. Its flagship product is speech-to-text, with models that transcribe both pre-recorded audio and real-time streams at low latency and high accuracy, supporting many languages, real-time code-switching between languages in a single stream, speaker diarization, formatting, and keyterm prompting that lets a customer add domain vocabulary instantly without retraining a model. It also offers a text-to-speech model tuned for the low latency that voice agents need, and a Voice Agent API that wires speech-to-text, a language model, and text-to-speech into one managed real-time pipeline so developers can build conversational phone and voice experiences without stitching the pieces together. Deepgram can run in its cloud or be deployed on-premises or in a customer's private cloud for data-sensitive workloads. It is used across call centers, media transcription, meeting tools, and the growing market of voice agents, and it publishes benchmarks positioning its accuracy and latency against other managed APIs. Deepgram is sold with usage-based pricing per hour of audio or per character, with free credits to start and volume and enterprise agreements for higher usage.