From Joe’s workshop
Every AI tool I’ve built (so far) 🤖
Here’s the deal. Over the last couple of years I got a little obsessed. Every time I caught myself doing something repetitive with AI (transcribing a video, cleaning up a prompt, pulling text off my screen) I’d stop and think, “Why am I doing this by hand?”
So I’d build a tool. Then another. Then another. This page is the whole pile, organized so you can actually find what you need.
Most of these run right on your Windows machine with AutoHotkey. A few use the cloud, a few run AI locally so nothing leaves your computer. The ones with a link (you’ll see a little ↗ next to the name) are ones you can grab for yourself. The rest are experiments, some half-baked, some just for me. That’s the fun of it. Time is the one thing you can’t buy more of, and every one of these gives a little of it back.
What AI tools can you build with AutoHotkey?
Way more than you’d think! This page lists 50 I’ve built: transcription (Parakeet, Transcriber), text to speech and voice cloning, writing helpers (ScoreMyCopy, AI Prompt Assistant), smart file renaming and search, OCR, background removal, video subtitles, and tools that help write AutoHotkey itself (AI Error Assist, V1 to V2 Converter).
Do these AI tools run locally?
Some do. They all run on your Windows PC with AutoHotkey, but the AI part varies. A few send work to the cloud, and a few run the AI locally (like Parakeet and Transcriber (Local LLM)), so nothing ever leaves your computer. Look for the ↗ next to a name: those are the ones you can download.
How AI Built a Page Listing All My AI Tools
Have a look around. If one of these solves a problem you deal with every day, that’s a win. 🚀
🎙️ Turning talk into text
I talk a LOT faster than I type. These take audio and video and hand you clean text, from live dictation to “transcribe this whole file and summarize it for me.”
AI Audio Transcribe ↗
Record on the fly or drop in a file, and it turns your speech into text with Whisper. Simple, fast, and it lives in your tray ready to go.
AI Audio Transcribe (WinRT) ↗
Same idea, but it grabs the audio straight from Windows instead of FFmpeg. One less moving part on your machine.
StreamSpeechToText ↗
Real-time dictation that types the words right where your cursor is as you speak. I run it with a foot pedal, which honestly feels like a superpower.
Transcribe and Summarize ↗
Point it at a recording and you get the transcript AND a summary. Right-click a file and go. Great for long meetings you’d rather not sit through twice.
Transcribe with OpenRouter
A little one-shot experiment: pick an audio file, send it through OpenRouter’s audio model, and copy back the text. Bare-bones on purpose.
Transcriber ↗
My workhorse. Transcribes, summarizes, and even whips up a YouTube thumbnail from your media. Right-click and Send-To make it effortless.
Transcriber (Local LLM) ↗
Same tool, but the summarizing happens on a local model. Nothing leaves your machine, and you’re not paying per call.
Parakeet
Drag a video onto it and it spits out subtitles right next to the file, all running locally on your GPU. No cloud, no waiting.
Transcribe Media (frozen build)
The rock-solid “known good” version of my full transcribe app, with speaker labels and filters. When I want something that just works, this is it.
Talk to AI ↗
Ask a question out loud and get an answer back from ChatGPT or Claude. It’s the closest thing to just having the AI in the room with you.
🔊 Making the computer talk back
Sometimes reading is slower than listening. These read text out loud, from plain clean voices to a clone of a real one with actual emotion.
Text to Speech ↗
Highlight text, hit a hotkey, and it reads it aloud with a natural AI voice. You can save the audio too. I use it to catch clunky writing by ear.
AI Clone Speak
Reads your selected text in a cloned voice through ElevenLabs. One hotkey, and there it is.
AI Clone Speak with Inflection ↗
The fun one. Drop tags like [excited] or [whispers] right into your text and the cloned voice actually acts it out. Way more human than flat narration.
✍️ Better writing, faster
Writing eats time. These help me score copy, match a client’s voice, and sharpen prompts before I send them.
ScoreMyCopy
Feed it a headline or product name and it grades it against the character limits and marketing basics: who’s it for, what’s the offer, where’s the reader in their journey. Takes the guesswork out.
Borrowing Writing Style
Keeps a profile of how each client writes so the AI can sound like them, not like a robot. Handy when you juggle a lot of voices.
AI Prompt Assistant ↗
Helps you build and polish a prompt, then fires it off to your AI of choice without leaving what you’re doing.
🗂️ Finding and fixing files
Point AI at a folder and just tell it what you want. Rename a mess of files, filter down to what matters, or search by meaning instead of exact words.
SmartFileRenamer
Describe how you want a folder of files named and the AI proposes the whole batch. No more renaming 200 files by hand. 😊
SmartFileSearch
Tell it what you’re after in plain English and it hands back only the files that match. Like having someone dig through the folder for you.
SearchOnConcept ↗
Drop in your docs, audio, and video, then ask a question. It finds the right file AND the exact page or timestamp where the answer lives. This one still amazes me.
Embeddings Doc Search
Semantic search over the AutoHotkey help docs. Type a concept, get the right help pages, even when you don’t know the exact command name.
🖼️ Working with images and video
Pixels in, results out: read text off the screen, make icons, strip backgrounds, chop up videos, burn in subtitles, and even generate animations.
AI OCR ↗
Drag a box around anything on your screen and it pulls the text out with AI vision, and it’ll translate it too. Text trapped in an image is one of my pet peeves, so this one gets used a lot.
IconCreator ↗
Type what you want, and it generates an image and saves it as a proper multi-size .ico. Every one of my tools needs an icon, so I built the tool that makes them. 😃
RemoveBackground
Drop an image, get it back with the background gone. Runs locally, so no API key and no upload. Fast and free.
Segmenter ↗
Breaks a video into logical chapters for you. Great for turning one long recording into bite-sized pieces.
Translate & Subtitle Video ↗
Generates translated subtitles and burns them right into the video. Reach a whole new audience without touching a video editor.
RemotionWrapper
Describe an animation in plain English (or pick a template) and it builds and renders the video for you. Motion graphics without the headache.
🛠️ Tools for building tools
A lot of what I do is AutoHotkey, so I built AI helpers for AHK itself: fixing errors, converting between versions, and inspecting what’s on screen.
AI Error Assist ↗
Your script throws an error, and instead of googling it, this sends the error and the code to an AI and hands you back a fix. Saves so much frustration.
AI Chat ↗
The same Error Assist idea, licensed and ready to hand out. Send an error or a chunk of code and get help back.
ErrorCapture ↗
Sits quietly in the background and catches your AHK errors the moment they happen, then logs them. Set it and forget it.
V1 to V2 Converter
Turns old AutoHotkey v1 code into v2. Meant to live inside your editor so upgrading old scripts stops being a chore.
V2 to V1 Converter
Goes the other way and flags anything tricky, with an optional AI pass to help with the parts that don’t map cleanly.
UIA Tree Inspector
Point at any element on screen and browse its accessibility tree, with an AI sidebar to help you write the automation. If you’ve ever fought with UIA, you know why this exists.
AutoHotkey Chat Rules
A quick hotstring that pastes my coding-style rules into an AI chat, so whatever it writes actually follows my conventions.
Figma (experiment)
An early design experiment. Not finished, but I keep it around because half of learning is the stuff that didn’t pan out.
⚡ Getting more out of the models
Working with the AI itself: comparing models, sharpening prompts, sorting out API keys, and running things right in the browser.
ModelBenchmark
Lists every OpenRouter model with its cost and context size, then runs a benchmark so you can see which one actually performs. Don’t guess which model to use. Measure. 😊
PromptImprover
Hit a hotkey and it rewrites your rough prompt into a much better one. A tiny bit of polish makes a HUGE difference in what you get back.
PromptStarting
Type your thought, and it opens a fresh Claude chat with a “keep it short and accurate” prefix already in place. Removes the friction of starting.
AI Hotkeys
AI-powered lookup and search shortcuts wrapped into one build. Answers a keystroke away.
ApiKeyIdentifier
Paste in a mystery API key and it tells you which service it belongs to. A small thing, but it’s saved me plenty of head-scratching.
GeminiNano
A web page that generates text with Chrome’s built-in on-device model. No server, no API key, no cost. The future is getting weird in a good way. 🤖
🤖 Assistants and automation
The bigger builds: voice interpreters, a live helper for sales calls, and command-line harnesses that drive other apps for you.
AI Interpreter
Speak, and it transcribes, runs your words through the AI, and speaks the answer back. Three models working together so you can just talk.
HeroCallAssistant
Listens to a live call and quietly surfaces helpful suggestions in real time, with you in the loop to approve them. Like a smart co-pilot on the phone.
AHK CLI (MCP)
A command-line “ahk” that lets AI validate, run, find, kill, and inspect scripts. This is how I let AI actually drive my AutoHotkey work.
CLI-Anything
A shared skin that gives all my little app-driving harnesses (GIMP, Blender, Inkscape, Audacity) the same clean, colorful terminal feel.
Claude Skills Viewer ↗
A clean dark viewer for browsing all the Claude Code skills you’ve got installed. Handy once your collection grows.
📚 Notebooks and data
Front-ends over the stuff I keep coming back to: research notebooks, contacts, and my to-do list.
NotebookLM Desktop
A desktop front-end for Google’s NotebookLM: create notebooks, add sources, and grab the results without living in the browser.
CardSnap
Pulls all my contacts down to a local database and a simple spreadsheet, so I can text and email people in bulk right from the PC.
VoiceToDo
A simple desktop client for my to-do list that syncs with the phone app. Topics on the left, tasks on the right, check them off as you go.
That’s the whole workshop. Some of these took an afternoon, some took weeks, and a few started as “this’ll never work” and turned into things I use every single day.
One thing I’ve learned: you don’t need a giant plan to start automating. Pick the one repetitive task that annoys you most, and take a swing at it. Worst case, you learn something. Best case, you never do it by hand again. 🚀
Want to learn to build stuff like this yourself? Come check out the courses over at the-Automator.com/Discover. They come with a 200% money-back guarantee, so you’ve got nothing to lose. 😊
Happy automating,
Joe Glines
