Audience
AI safety researchers and developers evaluating or building speech-based conversational AI.
About Dessa
Dessa is Block's speech-to-speech artificial intelligence model for conversational voice research and applications. It takes audio input and produces audio output directly, without converting speech into text as an intermediate step. Trained from scratch on human conversation rather than text, the model supports full-duplex interaction, listening and speaking at the same time, and runs locally on a MacBook Pro with an M4 Max chip or better. It captures conversational cues such as emotion, sarcasm, laughter, pauses, and turn-taking that can be lost in text-based pipelines. Dessa is intended for AI safety researchers and developers studying voice-model behavior, privacy, and real-time interaction. It is a speech model for research and development, not a general-purpose business application or hosted voice-agent platform.
Other Popular Alternatives & Related Software
GPT-Live-1.1
GPT-Live-1.1 is OpenAI’s full-duplex voice model for real-time conversational use in ChatGPT Voice. It listens and speaks continuously, so it can follow pauses, interruptions, and shifts in speaking pace rather than waiting for a clearly completed turn. In ChatGPT, Live conversations can combine voice with text and images, and may use web search, memory, and connected apps according to the account, plan, and workspace settings. Access is provided through ChatGPT experiences on the web and supported mobile apps, with availability and usage limits determined by plan. The model is intended for individuals and organizations that want spoken assistance for questions, brainstorming, information lookup, and everyday tasks. GPT-Live-1.1 succeeds GPT-Live-1 in OpenAI’s voice-model family.
Learn more about GPT-Live-1.1
GPT-Live-1
GPT-Live-1 is one of the two new GPT-Live voice models rolling out to ChatGPT users globally, built to make talking with AI feel much more like having a real conversation. It is powered by a full-duplex architecture, so it can listen and speak at the same time instead of waiting for one rigid turn to end before the next begins. During conversations, GPT-Live-1 can show it is paying attention with short acknowledgments, engage in quick back-and-forth, pause when the user needs a moment to think, or stay quiet when asked to listen. It continuously processes input while generating output, allowing the model to decide many times per second whether to speak, keep listening, pause, interrupt, or invoke a tool. GPT-Live-1 also separates natural interaction from deeper work: when a question requires web search, reasoning, or more agentic capabilities, it can delegate the task to a frontier model behind the scenes and bring the result back when it is ready.
Learn more about GPT-Live-1
GPT-Live-1.1 Quip
GPT-Live-1.1 Quip is an advanced voice AI model designed to make conversations with artificial intelligence feel more natural and intelligent. As the successor to GPT-Live-1, it delivers full-duplex voice interaction, allowing users to speak and receive responses continuously instead of waiting for clearly defined conversational turns. The model can follow pauses, interruptions, and changes in speaking pace, deciding in real time whether to respond or continue listening. This creates a more fluid, responsive experience that closely resembles natural human conversation. GPT-Live-1.1 Quip supports voice interactions across ChatGPT, Codex, and Dots, providing conversational capabilities for different AI-assisted workflows. The model incorporates safety training, voice-specific evaluations, and system-level safeguards designed to promote responsible interactions. Its training uses diverse datasets with filtering processes to maintain data quality.
Learn more about GPT-Live-1.1 Quip
GPT-Realtime-2.1
GPT-Realtime-2.1 is OpenAI’s reasoning model with tool use for low-latency voice agents and complex speech-to-speech workflows. It updates GPT-Realtime-2 with improved alphanumeric recognition, silence and noise handling, and interruption behavior, helping applications understand spoken code, manage imperfect audio, and respond more naturally when users pause or talk over the agent. Developers can configure reasoning effort to balance deeper thinking against latency and output usage, while strong instruction following helps the model stay aligned with a defined role, tone, and workflow. It accepts and produces both audio and text, can take images as input, and supports function calling so an agent can retrieve information or perform actions during a conversation. The model has a 128,000-token context window, supports up to 32,000 output tokens, and includes reasoning-token support for extended interactions.
Learn more about GPT-Realtime-2.1
Integrations
No integrations listed.
Company Information
Block
Founded: 2009
United States
block.xyz/dessa
Videos and Screen Captures
Dark application screen with a pale circular abstract icon centered above the transcript text, “Human winter is coming for sure but you.”
Other Useful Business Software
$300 Free Credits to Build on Google Cloud
Put your $300 in credit toward real workloads, then keep building with free monthly usage for 20+ products. No commitment and no charge until you upgrade.
Product Details
Platforms Supported
Mac
On-Premises
Dessa Frequently Asked Questions
Q: What kinds of users and organization types does Dessa work with?
Dessa works with these users and organization types: Mid Size Business, Small Business, Enterprise, Freelance, Nonprofit, Government, and Startup.
Q: What languages does Dessa support in their product?
Dessa supports these languages: English.
