Deepgram
A technology company specializing in speech recognition and voice AI. Its core offerings center on streaming real-time transcription with low latency and enterprise-grade custom model support, serving organizations that require scalable, high-performance speech-to-text infrastructure.
Core Facts
Timeline (last 90 days)
Deepgram announced the acquisition of real-time audio and video API platform Daily.
By acquiring Daily to fill in real-time transport and orchestration capabilities, Deepgram is laying groundwork in advance for the demand for voice agent applications.
Whether Pipecat's vendor neutrality is maintained after the acquisition depends on whether new component integration must go through Deepgram review, whether default configurations favor its own services, and whether maintenance of non-Deepgram adapters is marginalized.
If Pipecat is closed-sourced or deeply tied to Deepgram after the acquisition, the community will likely drift away.
Deepgram has long accumulated experience in the speech recognition (ASR) track, providing underlying transcription capabilities for a large number of voice applications.
Deepgram officially and explicitly committed to maintaining Pipecat's open-source nature.
Deepgram gives new users $200 in credit (about 46,000 minutes), after which payment is usage-based.
AssemblyAI provides an official Node SDK that supports uploading audio files to obtain transcript text and webhook callbacks; Deepgram's Node SDK supports real-time streaming transcription; Google Cloud's Node client is complex to configure and requires a service account key and project settings.
Deepgram's prepaid package costs about $0.0043 per minute, and AssemblyAI offers a free quota of 100 minutes per month.
Deepgram engineers found that audio feature extraction and attention computation are the main performance hotspots of Neuron kernels on speech models.
16 more timeline events
All Facts (20)
BYOK (Bring Your Own Key) mode allows developers to use their own API quotas purchased from OpenAI, Anthropic, or third-party aggregation platforms, sharing the same set of model resources across multiple IDEs.
75%Unverified在实时语音AI领域，OpenAI、ElevenLabs、Deepgram等公司都在布局相关产品
95%UnverifiedDeepgram在STT层以低延迟和高准确率著称
80%UnverifiedDeepgram以低成本高精度的ASR API占据开发者市场
75%UnverifiedBy acquiring Daily to fill in real-time transport and orchestration capabilities, Deepgram is laying groundwork in advance for the demand for voice agent applications.
50%UnverifiedDeepgram officially and explicitly committed to maintaining Pipecat's open-source nature.
50%UnverifiedDeepgram has long accumulated experience in the speech recognition (ASR) track, providing underlying transcription capabilities for a large number of voice applications.
50%UnverifiedIf Pipecat is closed-sourced or deeply tied to Deepgram after the acquisition, the community will likely drift away.
50%UnverifiedWhether Pipecat's vendor neutrality is maintained after the acquisition depends on whether new component integration must go through Deepgram review, whether default configurations favor its own services, and whether maintenance of non-Deepgram adapters is marginalized.
50%UnverifiedAssemblyAI provides an official Node SDK that supports uploading audio files to obtain transcript text and webhook callbacks; Deepgram's Node SDK supports real-time streaming transcription; Google Cloud's Node client is complex to configure and requires a service account key and project settings.
50%UnverifiedDeepgram's prepaid package costs about $0.0043 per minute, and AssemblyAI offers a free quota of 100 minutes per month.
50%UnverifiedDeepgram gives new users $200 in credit (about 46,000 minutes), after which payment is usage-based.
50%UnverifiedAccording to official Deepgram and AWS blog posts, after migrating to Trainium2, Nova-3's batch transcription throughput increased significantly and the cost per unit of audio processed dropped noticeably, but no precise benchmark numbers were published.
50%UnverifiedDeepgram engineers found that audio feature extraction and attention computation are the main performance hotspots of Neuron kernels on speech models.
50%UnverifiedAfter multiple rounds of kernel tuning, per-chip inference throughput improved substantially over the initial ported version, but the specific numbers were not disclosed.
50%UnverifiedDeepgram built a lightweight data-parallel runtime in Rust that splits audio files, distributes them to multiple Trainium2 instances, and aggregates the results, with support for dynamic batching.
50%UnverifiedNova-3 is Deepgram's flagship speech recognition model, supporting multilingual, high-accuracy, low-latency real-time transcription.
50%UnverifiedAfter the patient describes symptoms by voice, the audio is converted to text by Deepgram, and the agent executes a 5-hop Cypher query that completes diagnosis, contraindication checking, hospital selection, doctor assignment, and bed location in a single pass.
50%UnverifiedDeepgram removed several limited-time transitional prices (such as $0.041/min through 9/14 Then $0.059/min) and added multiple per-minute and per-character billing options.
50%UnverifiedThe developer uses Deepgram as the speech recognition engine, and the entire system supports more than 40 languages.
50%