Modulate

Most voice AI hears what was said. Velma hears what was meant. Tone. Intent. Emotion. Deception. The standard stack - STT to transcribe, LLM to analyse - throws all of it away before analysis even begins. The signal was lost the moment audio became text... (🧵)

@whuffman Don't settle for half-baked deepfake detection solutions. Explore how Modulate is doing it here: modulate.ai/blog/deepfake-…

The people on the other end of deepfake calls aren’t amateurs. They’re professionals. They study patterns, refine tactics over years, and scale - from small groups to nation-state operations. That’s the reality most people are up against In this clip, our CTO @whuffman explains why protecting people from these threats requires more than basic detection: it requires expert systems built to recognize how attacks actually work. That’s part of what we’re building at @modulate_ai Because individuals shouldn’t have to be fraud experts to stay safe 😶

Stop missing the signals that matter. See what your calls are really saying: modulate.ai/lp/velma-enter…

Velma is the world's first Ensemble Listening Model. 100+ AI models working directly on raw audio. Capturing the full signal. In real time, across every call. ✅40M+ business problems caught. ✅564M+ hours of conversation improved. ✅Fortune 500 deployed.

Most voice AI hears what was said. Velma hears what was meant. Tone. Intent. Emotion. Deception. The standard stack - STT to transcribe, LLM to analyse - throws all of it away before analysis even begins. The signal was lost the moment audio became text... (🧵)

if you picked deepgram for accuracy, look at this comparison. if you picked it for price, also look at this comparison.

Want to see how this works? Read more here: modulate.ai/press-releases…

Most voice AI tries to do everything with one model. That’s the problem. Here’s our approach: Not one giant model trying to guess everything -but a one of it's kind model, built one an ensemble architecture, designed to understand speech by breaking it into its core signals and analyzing each one properly ✅ In this clip, our CTO @whuffman walks through what makes Modulate’s approach different - and why it leads to something the industry rarely achieves at the same time: Higher accuracy. Lower latency. Real efficiency. Because real voice intelligence isn’t monolithic. It’s composable.

AI regulation is solving the wrong problem. Right now, most policies are built around generative AI: models like ChatGPT that create content (and yes, can hallucinate) But that’s only half the picture. There’s another category: analytic AI. Systems designed to understand what’s happening and return fixed, verifiable answers - no guessing, no hallucinations. In this clip, our CEO @mpappas74 breaks down why treating both the same is a mistake - and how current regulations are unintentionally slowing down tools that don’t carry the same risks. At Modulate, this distinction is core to how we build. Because not all AI should be regulated like it makes things up 👀

Voice chat changed gaming. It also made toxicity scalable 👀 In @CallofDuty: Warzone, real-time voice moderation powered by Modulate helped reduce severe toxicity exposure while improving player retention and engagement. This case study breaks down what happens when moderation moves from reactive reporting to proactive intervention - at global scale. Read the full story here: bit.ly/4eBLKdS @activision

The AI playbook says: more data, more compute, bigger models. We don’t buy it. At Modulate, this is how we think + how we build. In this clip, our CEO @mpappas74 breaks down why focused data + real insight beats brute force. We’re a team of ~40, and that approach has led to: - Transcription models outperforming @OpenAI on accuracy - Deepfake detection models topping the @huggingface speech arena leaderboard Not by hoovering the internet. By using the right data. Because better > bigger.

@hackernoon Full coverage by @hackernoon here: hackernoon.com/meet-modulate-…

We’re @hackernoon Company of the Week Voice AI breaks down when things get real: messy audio, emotion, overlap, intent. So we built Velma - the first Ensemble Listening Model, trained on 550M hours of real-world audio, designed to understand speech as it actually happens (not sanitized benchmarks). And ToxMod - real-time voice moderation that detects how something is said, not just the words. This isn’t research. It’s deployed at scale across Fortune 500 platforms today 🌐 Voice is the hardest problem in AI. It’s also the most human. We’re building the infrastructure to make it work -safely, accurately, in real time.

@jonashernlund @sama Exactly- this is the shift. We built Velma for it. sub-second streaming latency 25× better cost-performance Edge + real-time isn’t a constraint. it’s the product.

@sama Voice flips the cost model from per-token to per-minute. Latency budget is sub-200ms or it feels wrong, and most of that goes to audio not the model. Edge inference gets load-bearing fast. Whoever owns the local STT/TTS layer captures the user.

pretty excited for voice models to get great its interesting to watch how people are already starting to change the way they interface with AI

we’ve already built the best one @sama @modulate's Velma is a new class of voice AI: an Ensemble Listening Model. - #1 on Hugging Face (deepfake detection, 1.1% EER) - #1 in accuracy and cost-efficiency across benchmarks - 25× better cost-performance vs foundation models already powering production pipelines at Fortune 500 companies. worth a look: modulate.ai

🎙️Voice AI, explained like you’re 5 In this clip, our CTO @whuffman breaks down how voice AI picks up on tone, emotion, accent, and rhythm - then puts it all together to understand what’s really going on. Because “I’m fine” can mean a lot of different things. 👀 That’s the layer we’re building at @modulate_ai

Wrapping up two great days at @_odsc Thanks to everyone who stopped by the Modulate booth, asked sharp questions, and put our demos to the test. Until next time 👋

Hosting a deepfake happy hour at #BOSTechWeek this May Real vs synthetic audio Most people get it wrong Come try 👀 May 28, Boston RSVP: bit.ly/4ucWio8 @Techweek_

@_odsc lets go! 🙌

At ODSC AI East 2026, we’re proud to feature Modulate, a voice intelligence company building technology to make voice interactions safer, more inclusive, and more scalable. 📍 Meet Modulate at Booth #13 at ODSC AI East hubs.li/Q04dchfx0