Supercharging My Workflow with Superwhisper: A Deep Dive into AI-Enhanced Productivity
A comprehensive guide to my Superwhisper setup, including custom prompts, mode configurations, and integration with other AI tools for maximum productivity.
I'm always looking for ways to enhance my productivity workflow, and I've found that the right combination of AI tools can be a game-changer. I included Superwhisper in my AI Toolbox post, but I wanted to do a deeper dive into how I use it. Today, I want to do a deep dive into how I use Superwhisper, along with some other AI tools, to supercharge my daily work routine.
Why Superwhisper?
Superwhisper isn't just another speech-to-text tool - it's a powerful platform that can transform spoken words into various formats through custom modes and prompts. I've invested in the lifetime license because I knew I'd be "abusing it a bit" (in a good way!) with my extensive use of its features.
My Complete Setup
Let's break down my configuration in detail:
Voice Model Configuration
I use Nova 2 (Cloud) across all my modes, which provides excellent transcription accuracy while maintaining quick processing times. If you're espicially concerned about privacy, I recommend using the Ultra voice model. That said, the author of Superwhisper, Neil Chudleigh, has been updating regularly and often posts about improvements to the models. I highly recommend testing your setup with different models and combinations. You'll find that some prompts work better with certain models, some models perform better than others, and some combinations of models and prompts will yield better results than others. Once you get it dialed in, you'll find that Superwhisper is a game-changer for your productivity.
Language Models
Here's my current setup for models that I use in conjunction with the voice model:
- Claude 3.5 Sonnet: My primary workhorse for document and email processing
- GPT-4o: Specifically for certain message types and meeting summaries
- Both are cloud-based models, ensuring I always have access to the latest capabilities
Power of the Default Mode
The default mode is a powerful tool that can be used for a variety of purposes. I use this all the time for quickly dumping my thoughts. This mode simply transcribes my speech to text. I have found this mode to be particularly powerful while communicating with LLM chats and when working with tools like Cursor.ai. You input for those doesn't have to be perfect, can have mistakes, repeated words, spoken corrections, etc., and the tools will work with the output no problem. This is a great way to do brainstorming and get your thoughts out of your head and into a more structured format.
Custom Modes
I've set up several specialized modes, each with its own purpose:
1. Document Mode
The document mode is configured with:
- Language Model: Claude 3.5 Sonnet (Cloud)
- Voice Model: Nova 2 (Cloud)
- Language: English
In this mode, I use this specific prompt:
Please transform the following notes or descriptions into a structured markdown document.
Organize the content clearly, using appropriate markdown syntax for headings, lists,
and emphasis where necessary. Ensure the document is suitable for educational notes
or product documentation while preserving the text's original meaning, essence,
character, and informative nature.
This prompt ensures that my spoken thoughts are transformed into well-structured documentation.
2. Meeting Mode
This is one of my most sophisticated configurations. The meeting mode is set up with:
- Language Model: Claude 3.5 Sonnet (Cloud)
- Voice Model: Nova 2 (Cloud)
- Language: English
- Features enabled:
- Record from system audio
- Identify Speakers (crucial for meeting transcription)
The mode uses a comprehensive prompt that generates detailed, actionable meeting summaries:
# Meeting Summary Generator
## Meeting Transcript
{{TRANSCRIPT}}
## Instructions
You are an AI assistant tasked with creating a comprehensive, actionable meeting
summary. Follow these steps:
1. Carefully read the provided meeting transcript.
2. Extract and organize key information into these categories:
- Meeting details (title, date, attendees)
- Main topics discussed
- Decisions made
- Action items (with owners and deadlines if available)
- Other critical information
3. Document your analysis in the notes section, organizing the extracted information
into clear categories.
4. Create your summary following these markdown formatting rules:
- Use ATX-style headers (# with space after)
- Indent all lists and content under headers with 2 spaces for each level
- Indent content under list items by 4 spaces (aligned with text above)
- Use blank lines before and after headers
- Use blank lines before and after lists
- Use hyphens (-) for unordered lists
- Maintain line length under 80 characters
- Use asterisks (*) for emphasis
- Use proper spacing after list markers
- Use task list format for action items (- [ ])
Format your summary using this structure:
```markdown
# Meeting Summary: [Meeting Name] ([Date])
## Attendees
- [Name 1]
- [Name 2]
- [Name 3]
--- Begin Meeting Notes ---
- Key Topics Identified:
- Topic 1
- Topic 2
- Decisions Extracted:
- Decision 1
- Decision 2
- Action Items Noted:
- Owner 1: Action
- Owner 2: Action
- Additional Context:
- Context point 1
- Context point 2
--- End Meeting Notes ---
## Quick Recap
[2-3 sentences capturing the meeting's essence and main accomplishments]
## Key Decisions
- [Decision 1]
- [Decision 2]
- [Decision 3]
## Action Items
- [ ] [Owner]: [Action item with deadline if specified]
- [ ] [Owner]: [Action item with deadline if specified]
- [ ] [Owner]: [Action item with deadline if specified]
## Detailed Summary
### [Topic 1]
[Key points and relevant context]
- Sub-point 1
- Sub-point 2
- Additional detail
- Additional detail
### [Topic 2]
[Key points and relevant context]
### [Topic n]
[Key points and relevant context]
## Additional Notes
[Important context, follow-up meetings, or other critical information]
5. Important Guidelines:
- Title: Use specific meeting name from transcript. If none mentioned, create a
descriptive title based on content (e.g., "Q4 Product Strategy Meeting" rather
than "Team Meeting")
- Date: Use YYYY-MM-DD format if mentioned, otherwise use [Date Not Specified]
- Attendees: List all participants mentioned. If none stated, note "Attendees
not specified in transcript"
- Action Items: Always include owner and deadline (if mentioned), using task
list format
- Decisions: Write in past tense and include context if crucial for clarity
- Topics: Number them as needed (Topic 1, 2, ..., n)
6. Ensure your summary is:
- Clear and well-organized
- Comprehensive yet concise
- Action-oriented and practical
- True to the meeting's content and context
Please provide your summary following these formatting instructions.
The output follows a consistent structure:
- Meeting title and date
- Attendees list
- Key topics
- Decisions
- Action items
- Detailed summaries by topic
This format has proven incredibly effective for maintaining a clear record of all meetings and their outcomes.
3. Message Mode
For Slack and general messaging, I've configured a mode with these settings:
- Language Model: GPT-4o (Cloud)
- Voice Model: Nova 2 (Cloud)
- Language: English
This mode maintains natural conversation flow while ensuring clarity:
You are a transcription and text refinement assistant for text messaging and Slack apps. Your goal is to maintain the speaker's original voice and style with minimal adjustments. Follow these guidelines:
- Keep the original wording as much as possible.
- Correct obvious spelling mistakes.
- Add basic punctuation where needed (e.g., periods, commas, question marks).
- Break text into paragraphs for readability, but only if there are clear topic changes.
- Remove filler words (um, uh, like) if excessive or distracting.
- Eliminate repeated phrases unless used for emphasis.
- Only modify grammar or sentence structure if it seriously impedes understanding.
- Keep idiomatic expressions or colloquialisms as they are.
- Replace described emojis (e.g., "smiley face") with the actual emoji (😊). Keep any existing emojis as-is.
- For an empty message, reply with the original input.
Your goal is to enhance readability while preserving the speaker's voice, style, and emotion. Reply with the refined message only.
This mode is particularly useful for quick communications that need to maintain a casual tone while being clear and professional.
Integration with Other Tools
My workflow doesn't stop with Superwhisper. Here's how I integrate it with other tools:
Obsidian Integration
All my processed notes go into Obsidian, taking advantage of its markdown support and organizational features:
Work/
└── Meeting-Notes/
├── 2024-11-07 - Project Kickoff Meeting
├── 2024-11-06 - Weekly Team Sync
├── 2024-11-05 - Product Strategy Review
└── [More meeting notes...]
This structure keeps everything organized and easily searchable.
AI Model Integration
I've found that using different AI models for different purposes yields the best results:
- Superwhisper for initial transcription
- Claude or GPT-4 for deeper processing when needed
- Specific models for specialized tasks
Cost Management Strategy
Here's how I'm managing costs:
- Pay for Superwhisper lifetime license (worth it for heavy users)
- Let Superwhisper handle the basic AI processing costs
- Use dedicated Claude or ChatGPT subscriptions for more in-depth processing of raw transcripts when needed
Real-World Workflow Example
Here's how I use this setup in practice:
-
Capture Phase
- Speak thoughts, notes, or messages
- Use appropriate mode based on content type
-
Processing Phase
- Superwhisper handles initial transcription
- AI models apply formatting and structure
-
Storage Phase
- Paste into an Obsidian note in markdown format
- Review and make any necessary adjustments
- Organize in appropriate folders
- Tag for easy reference
-
Reference Phase
- Everything is searchable
- Consistent formatting makes review easy
- Action items are clearly marked
Tips for Setting Up Your Own Workflow
-
Start with Basic Modes
- Begin with document, message, and meeting modes
- Use the prompts provided above as starting points
- Adjust based on your needs
-
Optimize Your Prompts
- Be specific about formatting requirements
- Include examples in your prompts
- Test and refine based on results
-
Consider Your Output Needs
- Think about where your content will end up
- Ensure your formatting matches your destination system
- Plan for integration with other tools
Looking Forward
This setup continues to evolve as new AI capabilities become available. Current areas I'm exploring:
- Integration with more specialized AI models
- Automated action item tracking
- Enhanced meeting summary capabilities
The key is to build a system that's both powerful and flexible enough to adapt as your needs change and new tools become available.
This post is part of my ongoing series about productivity tools and workflows. If you're interested in more detailed breakdowns of specific aspects of this setup or other productivity tools I use, let me know! You can find me on X for direct requests.