Building a local video archive with Gemma 4 31B on a 2021 M1 Max MacBook Pro
By
asenna
Baker's choice. Dense with flavour, light on filler.
Summary
A detailed technical account of building a local, queryable video archive on a 2021 MacBook Pro M1 Max using Google's Gemma 4 31B model in LM Studio. The author describes their unique lifestyle split between wildlife filming in the Maasai Mara and software development, and their journey to index a year's worth of video footage locally. The article covers the technical challenges including 50GB of swap usage, four significant bugs encountered, and the practical lessons learned from running large language models on aging consumer hardware.
Key quotes
· 3 pulledI'm in the Maasai Mara about half the year, in three-month stretches. Animals out the front of the lodge, motorcycles, friends in the Maasai villages, kids who think a drone is the funniest thing they have ever seen.
The first half is a constant flood of footage from the iPhone, the DJI
While I slept, my 5-year-old MacBook ran Gemma 4 locally and indexed a year of video
You might also wanna read
PostHog plans to train its own AI models for proactive, self-driving product features
PostHog is entering a new phase focused on building proactive, self-driving AI-powered products. After launching popular AI features like th
posthog.com·4d agoPostHog plans to train its own AI models for proactive, self-driving product features
PostHog is entering a new phase focused on building proactive, self-driving AI-powered products. After launching popular AI features like th
posthog.com·4d ago
Models.dev: An open-source community database for AI model specifications and pricing
Models.dev is an open-source, community-contributed database that aggregates AI model specifications, pricing, and capabilities from various
github.com·9d ago
ModelHub: A macOS menu bar app for managing local LLMs across Ollama, MLX, and LM Studio
ModelHub is a native macOS menu bar app designed to streamline the workflow for developers working with local LLMs. It addresses the fragmen
Product Hunt·9d ago
LLMTest: Automated LLM Model Selection and Fallback Tool for Developers
LLMTest is a tool created by maker Tom to help developers and "vibe coders" automatically select the best LLM models for AI-powered features
Product Hunt·10d ago
DeepSeek-V4-Flash revives interest in LLM steering with local model capabilities
The article discusses LLM "steering" — manipulating model activations mid-flight to guide outputs — and highlights DeepSeek-V4-Flash as a br
seangoedecke.com·15d ago