Guides
How to choose speech-to-text software: dictation, meetings or an API
Pick speech-to-text software by job: dictation, meeting notes or a developer API. Compare offline use, free plans, speaker labels and pricing for six tools.
Choose speech-to-text software by the job first: dictation into whichever app has focus, transcripts of meetings, or a speech API that a developer builds into a product. The speech-to-text ranking keeps those three jobs apart, and this guide does the same. Accuracy depends on the audio, the language and the microphone, so the ranking relies on features and published terms rather than measured accuracy.
Match the tool to the job
Superwhisper dictates into any app on macOS, Windows, iOS and Android. Its free tier covers voice-to-text, and its Pro plan adds advanced modes at $8.49 a month.
Otter is aimed at meetings. It transcribes conversations, produces notes and offers iOS and Android apps, a Chrome extension and a web app. Its Basic plan is free forever, with a monthly limit on transcription minutes.
AssemblyAI and Deepgram are speech APIs for developers. Neither has an end-user app for dictation or meeting notes, so they fit software teams that want transcription inside their own product.
Whisper sits apart from these groups. It is an open-source speech model and Python package from OpenAI, released under the MIT license. It has no graphical interface, so it suits developers who are comfortable with Python and a local setup.
Setup effort differs as well. Whisper needs Python and a local install, while MacWhisper is a free download for Mac users who want their transcripts to stay on the machine.
Offline and local recognition
Offline use is the main privacy question. Whisper runs on your own hardware, so the audio does not have to go to a hosted service. MacWhisper transcribes audio, video and call recordings on a Mac, and its homepage says the data stays on the computer. It is available only for Mac. Superwhisper offers offline recognition alongside cloud recognition.
The ranking's FAQ says that the API and meeting tools in this list process audio through a vendor's service. Check that before you send interviews, client calls or any other sensitive audio to one of them.
Speaker labels and real-time output
Speaker labels are the main gap in the data. AssemblyAI lists speaker diarization, and the ranking marks speaker labels as not listed for the other tools. Check each vendor's documentation before you rely on speaker names in a transcript.
For live output, Deepgram lists real-time models for voice agents. The ranking does not list real-time output for the other tools, so test that feature on the vendor's pages if you need it.
Cost models
Consumer apps charge a monthly fee, and Superwhisper also offers a lifetime option. Otter's Pro plan is listed from $8.33 per user a month, and its Business plan from $19.99 per user a month. MacWhisper has a free version with a Pro option, and Whisper is free to use.
The APIs bill by audio time. AssemblyAI charges per hour of audio for each model, and its API can be tried for free. Deepgram charges per minute by model after a $200 starter credit. Because volume drives the cost of an API, estimate how many hours or minutes of audio you expect each month before you pick a model.
Our picks and what to do next
The speech-to-text ranking lists Superwhisper first, then Whisper, Otter, MacWhisper, AssemblyAI and Deepgram, in that order. Other tools sit between some of them. The positions reflect editorial judgments against the criteria, not a measured accuracy test.
- For dictation into any app, start with Superwhisper and use its free tier before you pay for Pro.
- For meetings, compare Otter's free monthly limit with the number of meetings you record.
- For transcripts of files and calls on a Mac, with the audio kept on the computer, MacWhisper is the option here.
- For a product that needs transcription, compare AssemblyAI and Deepgram on the model you need, and check speaker labels or real-time output on the vendor's pages.
The ranking also includes Wispr Flow (85.0), a dictation app that types into any app on Mac, Windows, iOS and Android and learns names and jargon. It has a free plan for individuals, while its Notetaker is limited to Mac and Windows and team features need paid plans.
If you plan to run models on your own computer more broadly, the local LLM runners guide covers that side of the work.
Compare head to head
Mentioned in this article
SuperwhisperVoice-to-text app that dictates into any app on desktop and mobile
WhisperOpenAI's open-source speech recognition model and Python package
OtterMeeting transcription and notes with apps and a Chrome extension
MacWhisperMac app that transcribes audio, video and calls on the device
AssemblyAISpeech-to-text and audio understanding models through an API
DeepgramSpeech-to-text, text-to-speech and voice agent APIs for developers
Related lessons
Learning Center
1:20How to compare products fairly: what to keep in mindDefine the job, cap the must-haves, compare like with like, count total cost over one period and weigh the criteria. Includes a worked example.
1:20How to run a software trial that tells you somethingRun a software trial that answers your question. Set criteria first, use real files, test import, export and sharing, then decide before the end date.Keep reading
All articles- Guides · 3 min readHow to choose an AI writing assistant: checks, rewrites or draftingChoose an AI writing assistant by task: proofreading, rewriting, long-form editing or drafting. Compare free plans, language coverage, team tiers and prices.
- Guides · 5 min readHow to compare AI chatbots without chasing leaderboardsA practical method for comparing AI chatbots by your own tasks, privacy settings, plan pricing, file handling and reliability, with rankings as a starting point.
- Roundups · 5 min readBest AI coding assistants compared by how they workCompare AI coding assistants by how they work, editor-integrated, terminal agent or open-source, with their pricing models and how much of the source you can read.
