Today I watched this YouTube, and I immediately gave it a try. It’s already transcribed and summarized 5 videos, I’m amazed, but I also have to admit, it failed on one longer video (76min recording).

This article is basically just a written form of what’s mentioned in that video - because I found the approach so practical that I had to document it for myself (and to share it easily with someone).

Instructions

Create a new Mode. Give it a name, for example “YT Summarizer”. Next step is you select custom mode. For the model I’ve currently chosen GPT-4.1 mini.

Here is the prompt:

You're given a raw transcription of a technical talk from a software development conference on YouTube.

Your task is to **convert it into well-structured, easy-to-read notes** that can be used as a reference or recap.

Follow these formatting and content guidelines:

---

## 📝 Format Instructions

- Use **clear section headings** to organize content (e.g., “About the Speaker”, “Key Concepts”, “Tool Demos”, “Action Items”).
- **Bullet points** only — no long paragraphs.
- Keep the tone casual and accessible — think dev blog or conference recap.
- Highlight key info like:
  - 🗓️ Dates
  - 🧑 Speaker names
  - ✅ Action items
  - ❓ Questions raised

---

## 📌 Content Goals

- Extract **core ideas**, tools mentioned, tips shared, and any code or commands worth noting.
- If there are **tasks, suggestions, or next steps**, list them clearly under an “Action Items” section.
- If **attendees ask questions**, collect and list them under a “Questions Raised” section.
- Avoid fluff. Capture the **essence** of the talk.

---

## 📚 Output Example

Use this structure as a rough reference:

```markdown
### Talk Title: “So You Think You Know Git”

🎥 [YouTube link](https://www.youtube.com/watch?v=aolI_Rz0ZqY)

---

#### 🔧 Git Tricks & Features

- **`git blame -wCCC`** – Blame smarter, skip whitespace, follow moved code.
- **`git maintenance start`** – Enables background cleanup & faster ops.
- **Force Push Safely** – Use `--force-with-lease` to avoid overwriting teammates' work.

---

#### 🆕 GitHub Tips

- Enable **auto-merge** after CI passes.
- Use **merge policies** to enforce rebase/squash/merge styles.
- Fetch any PR via its **ref** without adding remotes.

---

#### 🛠️ Tools & Demos

- **GitButler**: Virtual branching tool to multitask across features.
  - Drag & drop changes across branches.
  - Keeps context switching minimal.

---

#### ✅ Action Items

- Try `git maintenance start` on your main repos.
- Switch to `--force-with-lease` in team workflows.
- Explore signing commits with SSH keys.

---

#### ❓ Questions Raised

- Should our team be a “rebase” or “merge” shop?
- Could GitButler help our current workflow?

For the voice model I’m using Ultra, language is English, and then below there are Pro features where you have to toggle Record from system audio.

Then I go to the browser, open a YouTube video, select 2x speed and just let it run while the Super Whisper is listening. And once it’s done, I open Apple Notes and SuperWhisper pastes the summary into a new note.

Links & Resources