When you dictate into an app, two things are created: the audio you spoke and the text the engine produced from it. What happens to each one afterward depends entirely on how the tool works. Some apps record both and let you replay them. Some delete the audio immediately and keep only the text. Some store everything on a server. None of these approaches is universal, and the difference matters if you care where your voice ends up.
The three approaches to audio retention
Different dictation tools handle recordings fundamentally differently:
On-device, kept locally. The audio and transcript are both stored on your machine in the app's history. You can replay a dictation and see exactly what the engine heard. You own the storage and the app has no copy anywhere else. If you delete a history entry, the audio is gone from your device.
On-device, deleted immediately. Some dictation tools process the audio locally but delete it as soon as the transcript is created. You keep the text but never see the original audio. This saves storage and ensures that audio is never retained once it has been converted to words.
Cloud-based, stored on a server. The audio and transcript are sent to a server for processing and kept there in your account history. You can access them from any device. The server's retention policy determines how long they stay — some services delete them after a set period, some keep them indefinitely, and some offer a delete-on-download option.
Why retention matters
The choice between these approaches affects three things:
Privacy. If audio stays on your machine, only you can access it. If it reaches a server, the hosting company, potential attackers, and anyone with a legal demand for the data theoretically can. Even encrypted storage on a server is different from never uploading audio at all.
Recovery. If you delete a history entry on your local device, the audio is gone for good. If it is stored on a server, you may be able to recover it if the deletion was accidental — some services have a trash bin you can restore from. If you value the ability to retrieve old dictations, the history approach matters.
Storage. Audio files are large; a one-minute dictation is several megabytes. Keeping both audio and transcript for weeks' worth of dictations can consume significant space. Deleting audio immediately keeps only the small transcript. Cloud storage solves the space problem but at the cost of uploading and retaining.
How to know what a tool does
If a dictation app has a history view where you can see and play back old dictations, it is keeping the audio. If the history shows only text without a playback option, it is keeping only the transcript. If the history is on a separate server rather than a menu in the app, it is using cloud storage.
Most tools document this in their privacy policy or feature list. The key question to ask: "Does the audio reach a server?" If yes, your recordings are uploaded. If no, they stay local.
What this means for different use cases
For most people, the choice between on-device and cloud dictation is about offline capability and upload speed, not retention. But for sensitive conversations — dictating medical information, legal notes, or financial details — knowing where the audio goes is the most important question a dictation app can answer.