Official documentation

Privacy and local operation | Phorminx

Phorminx (Impossible G), 2026

Read the original

Read on October 21, 2026

What it is: Phorminx's own privacy page. It says speech recognition runs on the computer, that audio is held in memory and not uploaded, that formatting and assistant features use local models when chosen, and which connections are made by default or only when configured.
How much weight it carries: The maker's own statement of data handling, written as a list of connections. It describes the Windows app's behavior at the version it was read, and it is not an independent audit of the code.

Facts taken from this source

Phorminx's privacy page says audio is held in memory while it is transcribed and is not uploaded.
“Audio is held in memory while it is transcribed and is not uploaded.” (Speech recognition section)
Phorminx's privacy page says its audio is not uploaded.
“Audio is held in memory while it is transcribed and is not uploaded.” (Privacy page, 01 Speech recognition)
Phorminx's privacy page says dictation history has its own retention controls, and turning dictation history off changes future dictation retention.
“Saved sessions and dictation history have separate controls. Turning off dictation history changes future dictation retention.” (Retention section)
Phorminx's privacy page says feedback sent to phorminx.net is deleted automatically after one year.
“No audio, logs, or account details are attached and no IP address is stored with it. Feedback is deleted automatically after one year.” (Feedback section)
Phorminx's privacy page says its formatting uses local language models when the user chooses them.
“Formatting, live translation, the meeting assistant, and semantic search use local language models running on the same computer when you choose local models.” (Privacy page, 03 Local models)
Phorminx's privacy page says selected text goes to a remote assistant only when the user sets one up.
“If you set up a remote assistant provider or an HTTP action, the selected text and context go to the destination you configured, and only there.” (Privacy page, 04 Configured connections)
Phorminx's privacy page says speech recognition runs on the user's computer.
“Instant and accurate recognition run on your computer. Audio is held in memory while it is transcribed and is not uploaded.” (Privacy page, 01 Speech recognition)

Official documentation

Privacy and local operation | Phorminx

Phorminx (Impossible G), 2026

Read the original

Read on October 21, 2026

What it is: Phorminx's own privacy page. It says speech recognition runs on the computer, that audio is held in memory and not uploaded, that formatting and assistant features use local models when chosen, and which connections are made by default or only when configured.
How much weight it carries: The maker's own statement of data handling, written as a list of connections. It describes the Windows app's behavior at the version it was read, and it is not an independent audit of the code.

Facts taken from this source

Phorminx's privacy page says audio is held in memory while it is transcribed and is not uploaded.
“Audio is held in memory while it is transcribed and is not uploaded.” (Speech recognition section)
Phorminx's privacy page says its audio is not uploaded.
“Audio is held in memory while it is transcribed and is not uploaded.” (Privacy page, 01 Speech recognition)
Phorminx's privacy page says dictation history has its own retention controls, and turning dictation history off changes future dictation retention.
“Saved sessions and dictation history have separate controls. Turning off dictation history changes future dictation retention.” (Retention section)
Phorminx's privacy page says feedback sent to phorminx.net is deleted automatically after one year.
“No audio, logs, or account details are attached and no IP address is stored with it. Feedback is deleted automatically after one year.” (Feedback section)
Phorminx's privacy page says its formatting uses local language models when the user chooses them.
“Formatting, live translation, the meeting assistant, and semantic search use local language models running on the same computer when you choose local models.” (Privacy page, 03 Local models)
Phorminx's privacy page says selected text goes to a remote assistant only when the user sets one up.
“If you set up a remote assistant provider or an HTTP action, the selected text and context go to the destination you configured, and only there.” (Privacy page, 04 Configured connections)
Phorminx's privacy page says speech recognition runs on the user's computer.
“Instant and accurate recognition run on your computer. Audio is held in memory while it is transcribed and is not uploaded.” (Privacy page, 01 Speech recognition)