Chirp

Chirp represents the latest breakthrough in speech-to-text processing, developed by Google AI and integrated into Google Cloud's Speech API. This revolutionary model boasts 2 billion parameters and leverages self-supervised learning from millions of hours of audio and 28 billion text sentences across more than 100 languages. Chirp achieves a remarkable 98% speech recognition accuracy in English and a 300% relative improvement in several languages spoken by less than 10 million people.

General
Release date	2023
Author	Google AI
Type	Speech-to-Text

Standout Capabilities

Broad Language Support: Chirp caters to over 100 languages, ensuring top-notch speech recognition for a wide array of languages and accents.
Unparalleled Accuracy: With 98% speech recognition accuracy in English and notable enhancements in other languages, Chirp sets a new industry standard.
Massive Model Size: Chirp's 2-billion-parameter model outpaces previous speech models to deliver superior performance.
Innovative Training Approach: Chirp's encoder is initially trained with an enormous amount of unsupervised (unlabeled) audio data from 100+ languages, followed by fine-tuning for transcription in each specific language using smaller supervised datasets.

Start Building with Chirp

We have collected the best Chirp libraries and resources to help you get started and build state-of-the-art speech-to-text applications.

Chirp Libraries

A curated list of libraries and technologies to help you build great projects with Chirp.

Client Libraries

Chirp Boilerplates

Kickstart your development with a Chirp based boilerplate. Boilerplates is a great way to headstart when building your next project with Chirp.

Quickstart

Spparo Voice Agent

Real estate leads in Dubai message at all hours and often go cold before an agent can call them back. Spparo already qualifies these leads over WhatsApp. Spparo Voice Agent adds a voice layer: an AI agent that calls the lead, has a natural spoken conversation, and captures everything the agent needs. The agent listens to the lead in real time using AssemblyAI's Universal-Streaming (v3) speech-to-text over a WebSocket, with a server-minted temporary token so the API key never touches the browser. Each finalized turn is sent to an LLM through AssemblyAI's LLM Gateway, which understands the reply (in English, Hindi, or Hinglish), decides what information is still missing, and writes the agent's next question itself instead of following a fixed script. Across the call it extracts five things: name, buy or rent, preferred area, budget in AED, and timeline. A live transcript and an auto-filling lead summary appear on screen, and at the end the lead is scored hot or nurture. If the LLM Gateway is unavailable, a deterministic fallback keeps the call working so the demo never breaks. Roadmap: real outbound phone calls via Twilio, two-way natural voice through AssemblyAI's Voice Agent API, and direct integration with Spparo's WhatsApp flow so a voice call triggers automatically when a warm lead goes quiet, then pushes the result back into the CRM.

Fresno LUX: A Living City

Fresno LUX is a prototype for a living intelligence layer across an entire city. Fresno already contains enormous amounts of intelligence: residents know where problems exist, nonprofits know where help is needed, businesses know where opportunities exist, educators understand gaps in learning, researchers generate knowledge, and government maintains critical infrastructure and data. The problem is that much of this intelligence is fragmented across disconnected systems. Fresno LUX connects those pieces. The platform creates a shared city graph where people, organizations, resources, needs, projects, observations, experiments, and outcomes become connected nodes. Instead of simply collecting data, LUX continuously asks: What is happening? What do we know? What is missing? What can we test? What changed? A community problem can become an observable signal. That signal can become an experiment. The experiment produces evidence. Evidence changes the system's understanding. That updated understanding informs the next action. Observe → Connect → Understand → Act → Measure → Learn → Converge. The goal is to move Fresno from disconnected information toward collective intelligence—a system where knowledge can travel across boundaries and successful solutions can be discovered, tested, measured, and replicated. LUX is designed to make the invisible connections visible: the relationship between a neighborhood and its resources, between a student and educational opportunity, between a business and workforce capacity, between environmental conditions and community health, and between an idea and measurable real-world impact. This hackathon prototype demonstrates the beginning of a Living Fresno Graph: a city that can observe itself, learn from evidence, and coordinate toward better outcomes.

BlackBox - Video Captioning Agent

This project is built around the video captioning challenge: take a fixed set of short clips, thirty seconds to two minutes each, and write a caption for every one in four distinct voices. Rather than settling for one caption made safe enough for everyone, Blackbox writes four honest, independent takes on the same footage. For each clip it samples frames spread across the beginning, middle, and end, transcribes anything spoken in the audio, and then drafts all four captions at once rather than one after another, so a whole batch of clips can be worked through at the same time instead of queueing. It also checks its own work: every caption is graded against the clip for accuracy and tone before it's handed back, and anything that falls short gets quietly rewritten. Under the hood it leans on Google DeepMind's Gemma 4 as the first model to look at each clip, with Fireworks AI's minimax-m3 and qwen3p7-plus waiting as backups in case the primary model is busy or unavailable, Google Cloud Speech-to-Text for transcription, and a Dockerized Python pipeline .

Revv OS Powered by AutoSight

Revv is an enterprise-grade B2B SaaS platform designed to modernize the automotive retail experience. Today, dealerships struggle with static car listings and high-friction sales cycles. Revv solves this by utilizing AutoSight, our proprietary, autonomous multi-agent engine, to transform raw vehicle data into high-fidelity, cinematic 3D digital twins in under two minutes. The AutoSight Engine: At the core of the platform is a deterministic pipeline orchestrated by Google Cloud Workflows. This serverless state machine coordinates seven specialized AI agents powered by Gemini 3 Pro and Flash. The process moves from Ingestion (verifying real-world awards via Google Search Grounding) to Director-level decision-making, where the AI chooses between 2D imagery or interactive 3D WebGL "Tech Views." Using Imagen 3 for cinematic backgrounds and Gemini’s spatial vision for hotspot localization, AutoSight creates a narrative that explains complex technical specs through emotion and immersion. A Sovereign Infrastructure: To ensure true data sovereignty for dealerships, Revv is built on a distributed Vultr architecture. We utilize three dedicated Vultr Cloud Compute instances to host our frontend, our AI backend, and a self-hosted Supabase environment (PostgreSQL + pgvector). Massive images assets and neural audio tracks are offloaded to Vultr Object Storage, providing a scalable, secure, and high-performance environment that bypasses the limitations of shared public clouds. The Conversational Edge: For buyers, Revv offers the AutoSight Live Agent. Powered by the Gemini Multimodal Live API, this provides a real-time, low-latency speech-to-speech assistant. Backed by a Multimodal RAG system using Vertex AI Embeddings, the agent can "see" and query uploaded PDF manuals, extracting facts from complex tables and images to provide grounded answers. Shoppers can even book test drives mid-conversation via native tool-calling, which instantly triggers an interactive booking widget.