Wispr Flow

Wispr Flow Notetaker is an advanced AI-powered voice dictation and note-taking tool developed by Wispr AI, based in San Francisco. Launched in beta in late 2023 and made generally available in 2024, it leverages OpenAI's Whisper model for speech recognition, augmented by a proprietary language processing layer that removes filler words (e.g., "um," "like"), corrects grammar, applies punctuation, and learns user-specific names and jargon.

Architecture

Wispr Flow operates as a desktop or mobile app that captures audio via microphone and sends it to cloud servers for processing. Audio is not permanently stored locally, raising some privacy considerations. The processing pipeline is dual-stage: first, speech-to-text via Whisper; second, text refinement via a large language model (LLM) that adjusts tone and style based on the target application (email, chat, document).

Key Features

Real-time Dictation in Any App

The tool works via a global keyboard shortcut, enabling dictation into virtually any text field: text editors, messaging apps, email clients, and even search bars. This makes it a versatile productivity tool.

Filler Removal and Grammar Correction

It automatically deletes repetitive and unnecessary words, fixes grammatical errors, and inserts appropriate punctuation. The result is polished text ready for use without extensive manual editing.

Custom Dictionary and Vocabulary Learning

Users can add names, technical terms, or specialized vocabulary for the system to learn and improve accuracy over time. This is particularly useful for doctors, lawyers, and engineers.

Multi-language Support

The tool supports over 100 languages with automatic language detection, facilitating seamless switching during dictation.

Note and Meeting Transcription

Beyond short dictation, it can capture longer voice notes or meetings, with AI-generated summaries. This feature is limited on the free plan and expanded on paid plans.

Performance and Limitations

Tests show fast response times (under two seconds for short texts), but a stable internet connection is required. No offline mode exists. Arabic support is present but auto-refinement quality may be lower than for European languages due to less training data. No public API is available, limiting integration into custom workflows.