Qwen3.5 0.8B Does Real-Time Local Video Captioning

Melvin Vivas · X video post · 2026-03-26 · Open on X

Topics: LLM Fundamentals, Industry Trends & Job Market · Level: intermediate

Summary

A roughly 1GB Qwen3.5 0.8B model captions video in real time on a Mac Studio M2 Ultra. It takes under 1 second per frame and streams scene descriptions as the video plays. This shows how capable tiny local vision-language models have become.

Key points

• Model: Qwen3.5 0.8B, about 1GB.
• Hardware: Mac Studio M2 Ultra, running fully locally.
• Under 1s per frame; 269 frames captioned from a 3m49s video.
• Descriptions stream as the video plays, and you can pause anywhere to see its understanding of that scene.

Resources mentioned

• Qwen 3.5 · tool · qwen.ai · free
  Alibaba Qwen open model family with small on-device variants and toggleable reasoning.
  Also in: Workshop AI: Building Apps with Cloud and Local Agents (GLM 5, Qwen 3.5) (Melvin Vivas on X · notes), Qwen3.5 0.8B Real-Time Video Captioning on Mac Studio (Melvin Vivas on X · notes), Qwen 3.5 Is the Local Model That Holds Up in Claude Code (Melvin Vivas on X · notes), Free, Unlimited Agentic AI with Local Qwen 3.5 (Melvin Vivas on X · notes) and 7 more

Try this

• Build a local real-time video captioning app with a small vision-language model like Qwen3.5 0.8B.