DevelopingResearch

AssemblyAI's comparison splits Vapi, LiveKit, Retell and Pipecat by orchestration layer

Blockchain.News relayed AssemblyAI's Oct. 13 comparison, which argues Vapi, LiveKit, Retell and Pipecat sit at different layers of a voice-agent stack rather than competing on features.

Why it matters

The comparison is vendor-adjacent commentary from a speech-to-text provider about how four voice-agent products fit together, with no reported funding, contract, product change or regulatory consequence attached to it. Its significance is limited to developers choosing which layer of a voice-agent stack to build on.

Key facts

AssemblyAI published a comparison on Oct. 13, 2026 arguing that Vapi, LiveKit, Retell and Pipecat are different layers in a production voice-agent stack rather than direct substitutes.

Reported by Blockchain.News

AssemblyAI describes Retell as a managed, low-code phone-agent platform that emphasises visual builder workflows and built-in telephony.

Reported by Blockchain.News

AssemblyAI describes Vapi as an API-first orchestrator that exposes the conversation pipeline through code and API configuration and abstracts telephony and SIP setup.

Reported by Blockchain.News

AssemblyAI describes LiveKit as open-source WebRTC transport plus an agent framework that runs on WebRTC with an SFU and joins a room as a participant in an event-driven model.

Reported by Blockchain.News

AssemblyAI describes Pipecat as an open-source Python pipeline developers assemble themselves and says it is the only one of the four where every step is directly inspectable and modifiable.

Reported by Blockchain.News

The reported comparison states all four products depend on the underlying speech-to-text model.

Reported by Blockchain.News

The reported comparison says two of the four products are often used together rather than treated as rivals on a single feature axis, without naming which two in the evidence provided.

Reported by Blockchain.News

What happened

Blockchain.News reported on Oct. 13, 2026 that AssemblyAI published a comparison arguing Vapi, LiveKit, Retell and Pipecat are not direct substitutes but occupy different layers of a production voice-agent stack, with the real choice being how much of the pipeline a team wants to own. AssemblyAI places Retell as a managed, low-code phone-agent platform built around visual builder workflows and built-in telephony, while Vapi is described as an API-first orchestrator that exposes the conversation loop through code and API configuration. LiveKit is characterised as open-source WebRTC transport plus an agent framework using an event-driven model in which the agent joins a room as a participant, and Pipecat as an open-source Python pipeline that developers assemble themselves, with the source stating Pipecat is the only one of the four where every step is directly inspectable and modifiable. On transport, the report says Retell handles telephony and call transfer, Vapi abstracts telephony and SIP setup, LiveKit runs on WebRTC with an SFU, and the Pipecat portion of that transport comparison is cut off in the evidence. AssemblyAI also argues all four depend on the underlying speech-to-text model and that two of the products are often used together rather than treated as rivals. No other outlet in the evidence set covers the story, so none of the layer assignments or the pairing claim has independent confirmation here, and the evidence does not state which two products the article says are commonly combined.

How the story developed

  1. First report

    First report by Blockchain.News

    Vapi, LiveKit, Retell and Pipecat differ by orchestration layer

    Blockchain.News

  2. Update

    AssemblyAI publishes voice-agent stack comparison, relayed by Blockchain.News

    The comparison separates Vapi, LiveKit, Retell and Pipecat by orchestration, transport and transcription control, and was relayed by Blockchain.News.

    Blockchain.News

Affected entities

Original sources

PublisherReportRolePublished
Blockchain.News
Crypto media
Vapi, LiveKit, Retell and Pipecat differ by orchestration layerOriginal