Arjun Kava

👋 Hey Product Hunt + Twitter, I'm Arjun — co-founder of @VideoSDKLive. Support us here: producthunt.com/products/video… We just open-sourced something I'm really proud of: VideoSDK Real-Time AI Agents. If you're building voice agents — this infra will change everything. Let me explain 👇

Which means the same system can be used to build: • transcription pipelines • voice copilots • chat + voice agents • fully autonomous voice agents • realtime agents No rewrites. No separate stacks. videosdk.live/blog/introduci…

Introducing Prism: Agents v1 (VideoSDK) • Unified Pipeline • pipeline hooks - control any stage (audio, text, turns) • Hybrid Pipeline - Mix any stack - STT, LLM, TTS, knowledge base, actions • Compose anything - STT-only, LLM-only, full voice, realtime • Built-in observability - metrics, logs, traces • Avatar support

You don’t like the voice of your realtime model. But you can’t change it. Can’t fix pronunciation. Can’t clean transcripts before LLM. That’s where voice agents fail. So we rebuilt the pipeline.

@video_sdk x @Anam__ai is live.

Anam interactive avatars are now natively supported on @Video_SDK. Add real-time video avatars to VideoSDK agent pipelines. Under 10 lines of Python. Sub-second response times. Works with your existing STT, LLM, and TTS. 70% user preference over voice-only. CARA model leads all tested providers on visual quality, lip sync, and overall experience (avatarbenchmark.com). anam.ai/interactive-av… @SagarKava_ @Arjun_Kava

@bevenky Thanks, Venky. Big fan of what you've built at Plivo. 🫡

@Arjun_Kava This is the missing piece. Every team building voice agents has been duct-taping deploys together. CLI + Dashboard with version control and real-time debugging is exactly what the space needs. The "Vercel for voice agents" moment is here.

1/ Introducing AI Voice Agent Cloud Deployments via CLI + Dashboard ⚡️ Deploying voice agents shouldn’t feel complicated.

@operator_mode Cold start under 500 ms

@Arjun_Kava what's the cold start latency looking like for voice specifically? that's usually where cloud deployments get tricky

6/ If you’re building voice agents / real-time assistants / autonomous workflows — this is built for you. videosdk.live/blog/introduci…

5/ No complex infra setup. No DevOps overhead. Just ship AI agents.

8/ 🚀 One Platform. Any Voice AI model. Engineered for enterprise-grade reliability. Start building today 👉 dub.sh/uIsIJu9

7/ Supported models (and growing): - @GeminiApp - @SarvamAI - @DeepgramAI - @cartesia

1/ Announcing VideoSDK Inference: One Magic API for every Voice AI model 🎉