Summary
Highlights
The video introduces the concept of AI dictation tools, highlighting their ability to understand intent and format text for specific applications (email, text, Slack, code editor). It contrasts them with traditional speech-to-text, which often requires manual editing. The goal is to compare three popular AI dictation apps: Whisper Flow, Super Whisper, and Mac Whisper, against Apple's native speech-to-text.
The presenter outlines the pricing models for each tool. Whisper Flow offers a free plan with 2,000 words per week, and a $15/month (or $12/month annually) premium plan for unlimited words and pro features. Mac Whisper is a one-time purchase of $49 for Pro features, granting access to LLMs like ChatGPT and Claude, while a free version with smaller models is available. Super Whisper provides a 15-minute free trial for pro features, with monthly, annual, and lifetime options ranging from $8.50/month to $250.
Whisper Flow is praised for its clean UI and intuitive design, similar to a Mac app. It offers customizable hotkeys and gamified dashboard features like word streaks and words per minute. During testing, it accurately transcribed a simple sentence and maintained good punctuation. However, it failed to understand instructions to format text as a 'text message', treating it as dictation. The tool also didn't use exclamation points for vocal inflections. A significant point of concern is its constant internet connection and cloud-based processing, leading to high CPU usage and data privacy worries.
Super Whisper is presented as a more complex tool, allowing users to choose from various AI models with different speeds and accuracies. It offers features like text replacement and configurable modes for messages, emails, and documents. In testing, it performed similarly to Whisper Flow in transcribing sentences and ignoring filler words. Like Whisper Flow, it struggled to interpret formatting instructions, simply transcribing them as spoken. It also did not add exclamation points based on vocal tone. Super Whisper appears faster than Whisper Flow due to internal processing.
Mac Whisper is highlighted as the most comprehensive tool, capable of transcribing podcasts and recording meetings, with dictation currently in beta. It processes all data locally on the device, requiring users to download models. This local processing makes it very fast. In transcription tests, Mac Whisper performed well, similar to the other tools, even with filler words. It also failed to understand instructions for formatting, simply transcribing the user's commands. Exclamation points were not used based on vocal emphasis.
The video compares the AI dictation tools to Apple's built-in speech-to-text. Apple's tool performed surprisingly well, accurately transcribing sentences and ignoring filler words, putting it on par with the dedicated AI tools for basic dictation. One key difference observed is that Apple's tool displays text as it's being spoken, suggesting it doesn't perform the same level of AI reasoning or contextual understanding as the other tools.
A crucial test involved dictating an email. Whisper Flow was the only tool that automatically formatted the dictated text appropriately for an email, demonstrating a significant advantage in understanding context. Super Whisper and Mac Whisper failed to apply email formatting, presenting the text as a continuous block. The presenter concludes that while all tools are similar for basic dictation, Whisper Flow excels in contextual formatting. Integrating these tools into workflows can significantly increase productivity by allowing users to speak their thoughts directly.