AI agent that speaks during meetings

Welcome back.

Otter’s new voice-activated Meeting Agent now speaks up during calls, handles tasks, and remembers everything from past meetings.

But is it actually useful?

Let’s find out.

In today’s release:

1. Otter’s AI agent for meetings

2. New powerful AI image generator

3. DeepSeek updates its non-reasoning model

Otter’s AI agent can now talk in meetings

Otter has launched a new voice-activated Meeting Agent that can speak up during Zoom calls, answer questions, and complete tasks using your team’s past meeting data. It’s a major upgrade from Otter’s usual transcription tool, giving users real-time voice assistance.

Instead of just transcribing and summarizing, Otter’s Meeting Agent can now schedule meetings, draft emails, and respond to questions live, all through natural voice commands. It’s currently rolling out for Zoom, with Microsoft Teams and Google Meet support coming soon.

• Voice-activated: Ask questions or give tasks using natural speech.

• Context-aware: Pulls answers from your company’s past meeting data.

• Multi-platform: Rolling out on Zoom now; Teams and Google Meet next.

Otter is also releasing two more agents: a Sales Agent that coaches reps live on calls, and an SDR Agent that demos products on its own. Both are aimed at streamlining workflows and reducing manual effort during calls.

How can you use this to your advantage?

With Otter’s Meeting Agent, you can turn meetings into interactive work sessions, schedule follow-ups, and draft content without pausing the conversation. It’s almost like having a smart teammate that never forgets anything.

Reve Image 1.0: The new leader in AI image generation

Reve AI has launched Reve Image 1.0, a powerful AI image model focused on prompt accuracy, aesthetics, and text rendering. It’s already outperforming industry giants like Midjourney v6.1, Imagen 3, and Recraft V3.

The model can generate high-quality visuals from simple text and lets users modify existing images with natural language. It also excels at rendering readable text within images, a longtime weakness for most image AIs. Reve Image supports multi-character scenes, reference uploads, and style tweaks through intuitive settings.

• Benchmark leader: Ranked #1 by Artificial Analysis for image quality.

• Text rendering: Beats Midjourney and rivals Ideogram in clear, readable text.

• Customization tools: Style prompts, aspect ratios, and multi-image options.

• Prompt enhancer: Automatically adds visual detail to basic prompts.

Reve is still in preview, with no API or pricing announced yet. While it has minor issues with complex or fictional content, user feedback suggests it’s a major leap in creative AI tooling.

How can you use this to your advantage?

If you create branding, social content, or concept art, Reve lets you skip complex prompting and get clean, legible, on-style images fast. It’s especially valuable for visuals that need real text, consistent style, or multi-subject scenes.

DeepSeek updates its non-reasoning AI model

DeepSeek has pulled off a major open-source win, its new V3-0324 model just became the top-ranked non-reasoning AI on the Artificial Analysis Intelligence Index. It surpassed big names like Claude 3.7 Sonnet, Gemini 2.0 Pro, and Llama 3.3 70B in benchmark scores.

V3-0324 is designed for real-time performance, producing fast responses without lengthy “thinking” phases. It’s suitable for chatbots, customer service, and translation, apps where speed matters more than deep logic. While it still trails “reasoning” models like DeepSeek R1, this release pushes open-source tools much closer to parity.

• Leaderboard jump: Gained 7 points to top all non-reasoning models.

• Specs: 671B total params, 37B active, 128k context (64k via API).

• MIT Licensed: Free to use, modify, and integrate commercially.

• Limitations: Requires over 700GB of GPU memory—strictly enterprise-grade.

It’s a major milestone: the first time an open-weight model leads its category, showing just how far open AI has come. DeepSeek’s steady momentum has turned it into a serious contender against the world’s biggest AI labs.

How can you use this to your advantage?

If you run real-time applications like support bots or translation tools, V3-0324 delivers fast responses without sacrificing quality. And since it's MIT-licensed, you can build on it freely, if you’ve got the GPU power.

OTHER AI NEWS

ChatGPT Voice interrupts less often

OpenAI updated its voice assistant to reduce interruptions when users pause or breathe mid-sentence. All ChatGPT users can now enjoy smoother, more natural conversations with the AI.

A startup 'Paid' helps AI agents get paid like employees

Outreach founder Manny Medina launched Paid, a billing platform for AI agents that charges based on results, not usage or seats. Backed by €10M in pre-seed funding, it aims to help agentic startups set pricing, track ROI, and manage margins.

Australian authors outraged over Meta AI training on pirated books

Writers discovered their books were likely used without consent to train Meta’s AI using the pirated LibGen dataset. Authors like Holden Sheppard and Tracey Spicer called it a violation and are pushing for compensation and stronger AI copyright laws in Australia.

POPULAR AI TOOLS

1. Inferless → Deploy any machine learning models in minutes.

2. Supercut → Record your screen, share, look like a PRO.

3. IP Portal → Protect, license, and monetize your IP on your own terms.

4. AI Voices by Asyncflow v1.0 → Premium AI voice quality without the premium price tag.

5. Zapier MCP → Connect your AI to any app with Zapier MCP.

AND THAT’S A WRAP

Thank you for reading!

If you found this email useful, share it with a friend or colleague who also loves AI.

Also, drop me a follow on Twitter/X for more AI and tech updates.

I will talk to you soon!

Mike