AI Use Case: Create Tasks and Notes via Voice

Speech-to-Text for (almost) any app

I have hundreds of ideas every day—most of them utterly useless. However, one or two flashes of inspiration usually feel valuable. The problem is that I forget 99% of them, watching them disappear into the mysterious abyss from which they came. The same happens with tasks: when I get a delivery, I remember I need to email the building manager about fixing the door buzzer, but by the time I'm back at my desk, it has vanished from my mind.

The challenge of quickly capturing ideas and tasks is as old as human organization itself. With recent AI breakthroughs in Natural Language Processing and Large Language Models, this should be simple: speak an idea into your phone, let AI polish it, and find it in your preferred app. Yet in talking with busy knowledge workers, I discovered this isn't reality. Most apps either lack speech-to-text capture or implement it poorly. Anyone who uses ChatGPT or other AI apps knows what good transcription feels like—and standard OS-based voice transcription falls short (I'm looking at you, Apple). I needed something as reliable and forgiving as ChatGPT's and Gemini's voice dictation.

As a devoted user of Todoist and Notion, my goal was straightforward: dictate new tasks and notes with zero friction. Neither app currently offers built-in speech-to-text processing. After testing dozens of third-party solutions, I finally found one that works perfectly. While I discovered several standalone options, most tried to replace my existing tools rather than simply handling the capture process. This reflects a broader issue in AI products today—everyone wants to become your complete AI interface.

Solution Overview

1. Trigger the workflow

2. Record the audio task/note

3. Transcribe the audio with a modern speech-to-text model

4. Use AI to generate a short title for the task/note

5. Add the title and full transcription as a task or note to your favorite app

I’m keeping my solution simple so you can adapt it to your personal tech stack. Just swap out the components as needed, and pick the solution that matches your technical skill level and personal preference. Here are two specific solutions to get you started.

Option A: Using the SuperWhisper and ChatGPT app

SuperWhisper is one of my top discoveries this year. I frequently use it across various applications to dictate emails, messages, notes, and more. It works on both Mac and iPhone, providing flawless and truly seamless voice transcription as well as AI processing of that transcription. It uses all the latest AI and voice models, you can bring your own API key if you want, and even local models are available for offline use. The best thing: SuperWhisper offers a lifetime license - it isn’t yet another monthly subscription I need to add to my budget. SuperWhisper can do so much more; go check it out.

I assigned SuperWhisper to the action button on my iPhone. After launching the app, I hit record, speak my message, and the transcription occurs. Then, I tap "Share." In the share sheet, I select my custom "Create Task in Todoist" shortcut, which automatically sends the dictated task—complete with a concise title—to my Todoist Inbox for later processing.

Here's how the "Create Task in Todoist" custom shortcut works:

1. Receive Input from Share Sheet

   1. Store in new variable transcript

2. Create Text for the ChatGPT prompt:

   1. You are an expert in creating very meaningful task titles. You will receive transcribed voice input for a new task that you need to create a task title for. The task title needs to be short but very clear. Only generate the task title, nothing else. Don’t provide any options. Just one meaningful task title. Here is the transcript: transcript (add the previously created “transcript” variable here)

   2. Store in new variable prompt

3. Send the prompt to ChatGPT in a new chat using your preferred model (I use GPT-4.1-mini for its good balance of speed and accuracy).¹

   1. Insert the prompt variable as input.

   2. Uncheck both "Continuous chat" and "Show When Run" options.

   3. Save the response in a new variable called taskTitle.

4. Create a new task in Todoist, using taskTitle for the title and transcript for the description.

⬇️ Download the Apple Shortcut: Create Task in Todoist | Create Note in Notion

Option B: Use OpenAI APIs

I said it before, but because it’s so true, I’ll say it again: Everyone tries to be the frontend for your AI needs. However, you can bypass the middleman and just build your own frontend on top of the major AI APIs (i.e., OpenAI, Google, Anthropic, …). Thanks to vibe coding (→ bolt.new) and powerful automation tools (→ Apple Shortcuts), you don’t need to be a frontend engineer anymore to make this happen. You can just do it.

I created a custom iPhone shortcut that handles the entire process: it records audio, uses OpenAI's "transcriptions" API to convert speech to text, sends the transcript to the "responses" API (using GPT-4.1-nano for example) to generate a task/note title, and saves everything to my Notion "Inbox" database. You'll just need your OpenAI API key and Notion API key to get started.

⬇️ Download the Apple Shortcut: Transcribe Audio

Once downloaded, add the shortcut to your home screen or assign it to your action button.

This solution offers even more flexibility than the first one—any service with an API can be your destination for dictated content. Feel free to use my template and adapt it to your needs. If you make improvements, I'd love it if you shared them back with me.

Potential Next Steps

I wanted my initial solution to be as simple and flexible as possible. But there is definitely more that could be done to improve the experience:

• Use a voice command to trigger the voice recording (”Hey Siri, transcribe a new idea”).

• Select different target databases for Notion for different types of content.

• Use an AI model to extract project, due date and priority from the transcription for a more targeted save in Todoist.

Links

During my research, I came across many great apps. Maybe one of them is ideal for you:

• Tana: A full notes solution with great ideas (supertags!) and a dedicated (voice) capture app that can extract specific content from what you dictated. I really love that one, but I wasn’t ready to replace Notion yet.

• Voicenotes: A really slick voice note capture app that creates transcriptions, allows you to run prompts on those voice notes (e.g., extract to-do’s) and automatically syncs with your favorite tools (e.g., Notion via Webhook). The only reason I’m not using it is the monthly price. But I’m tempted…

• Funnel: A beautiful and simple app to capture all types of content quickly and send them to your favorite services (Notion, Todoist, and many more). The transcription is based on the OS-based feature (as far as I can tell), which is not great. Apart from this, a wonderful app with a lot of potential.

1. Using the ChatGPT app for automation presents one issue: the app must remain active in the background; otherwise, the shortcut fails. However, if you use it frequently, this shouldn't be a problem. Alternatively, you can use the OpenAI API as described in Option B.