The Best Local AI for Each Job, at Every Memory Size
Description
The best local AI model for each job, and how much memory each one needs: dictation, transcription, text to speech, image generation, reading documents, writing, coding and agents, on a Mac or PC with 8 to 128 GB. ๐ Free cheat sheet from this video, with every pick, its app, download size, memory and settings, plus which speech model covers which language: https://theaibridges.com/resources/best-local-ai-each-job-cheat-sheet?utm_source=youtube&utm_medium=description&utm_campaign=best-local-ai-each-job ๐ Local AI on Your Mac or PC, my guide, from picking the computer to your own private ChatGPT and agents, now with everything from this video as a new chapter ($149): https://theaibridges.com/learn/local-ai-mac?utm_source=youtube&utm_medium=description&utm_campaign=best-local-ai-each-job ๐ LLM Sizer, my free tool for what runs on your computer and how fast: https://theaibridges.com/tools/llm-sizer?utm_source=youtube&utm_medium=description&utm_campaign=best-local-ai-each-job Twelve everyday jobs, one free pick for each, and a demo of every one on my 128 GB M5 Max MacBook Pro. For each job I go through what people pay for today, the free local models, my pick, the settings that matter and the smallest machine it runs on. The models and apps in the tests: Parakeet, Whisper, Qwen3-ASR and NVIDIA's Nemotron for speech, Ming for cloning a voice, FLUX.3 klein in Draw Things for pictures, Qwen 3.5 9B, Qwen 3.8 27B and Gemma 4 in LM Studio for photos, questions, emails, documents, presentations and small tools, and Hermes Agent for agents. The last part shows what you can run with 16, 24, 32, 48 to 64 and 128 GB of memory. Videos mentioned: Qwen 3.8 against Claude Opus 4.6: https://www.youtube.com/watch?v=Qck0_OeFYWU Local image models on a Mac, tested in September: https://www.youtube.com/watch?v=igaKGikdlng What AI you can run at every memory size: https://www.youtube.com/watch?v=vwvT01tT5Ac ๐ฌ The newsletter (AI for business, no hype): https://theaibridges.substack.com Chapters: 00:00 The best local AI for each job 01:02 Memory and download size, explained 01:48 Job 1: Dictating into any app 06:05 Job 2: Transcribing recordings and calls 09:50 Job 3: Text to speech in your own voice 13:00 Job 4: Generating pictures 15:58 Job 5: Reading photos, receipts and scans 19:31 My Local AI guide 20:00 Job 6: Answering questions 25:58 Job 7: Writing emails 27:30 Job 8: Reading your documents 31:08 Job 9: Making a presentation 34:17 Job 10: Building a small tool 37:02 Job 11: Running an AI agent 41:07 Job 12: Building a real app 42:54 What runs at 16, 24, 32, 64 and 128 GB Connect: YouTube: https://www.youtube.com/@alexgrankin LinkedIn: https://www.linkedin.com/in/alex-grankin X: https://x.com/alex_grankin Site: https://theaibridges.com #LocalAI #LMStudio #LocalLLM
5:09I Built an Offline AI That Reads Your PDFs and the Pictures in Your SlidesExplain AI25 views
10:49Running AI Agents Locally = Safe...? Think AgainDaniel Jindoo 71.2K views
1:27:42Building an AI Plant Care App LIVE on Biela.devbiela dev126 views
2:39Solar Powered Raspberry Pi wildlife identifier #raspberrypi #ai #cameraLuke Ditria8.3K views
39:41my local, AI Voice Assistant (I replaced Alexa!!)NetworkChuck1.6M views
12:19Self-Hosted AI That's Actually UsefulTechno Tim246K views
10:09AI Agents, Clearly ExplainedJeff Su5.7M views
29:12Master Local AI in 29 minutes (LM studio + AnythingLLM)Keith AI110K views
26:00Local AI Agents In 26 MinutesTina Huang841K views
2:38:37The Ultimate Guide to Local AI and AI Agents (The Future is Here)Cole Medin286K views
3:39My Plant Can Talk! (Building an AI Plant Guardian with Raspberry Pi)Bondan Vitto42 views
16:34The Unbeatable Local AI Coding Workflow (Full 2026 Setup)Zen van Riel303K views
5:06How to Train an LLM on Your Own Data: Tips for BeginnersDecodo (formerly Smartproxy)87.6K views
26:01Your Local AI Is Not as Private as You ThinkKai605K views