Privacy policy

Information stored

Hearlog has no accounts or servers. Hearlog stores recordings, transcripts, summaries, translations, bookmarks and notes, group names, your vocabulary, your name, the summary format and your instructions, usage records, API keys, downloaded local models, call recording folder locations, lists of files in call recording folders (including file names, sizes, modification times, and files not imported into the app), the email address of your connected Google Drive account, and settings on this device. This information is sent outside the device only in the cases described under 'Information sent outside the device' below.

Hearlog also stores the addresses and usernames of connected WebDAV and SFTP servers; the endpoint, region, bucket, and access key ID of connected S3-compatible storage; and the email address of your connected OneDrive account on this device.

Hearlog also stores the addresses and model lists of connected model servers (Ollama and LM Studio) on this device.

Information sent outside the device

Hearlog sends content for each step that uses a Gemini model to the Google Gemini API using your API key, over HTTPS only. For transcription, Hearlog sends the recording file and your vocabulary. The original file name and your name are also sent only if the recording is a call recording or 'Show speaker names' is enabled. For summarization, Hearlog sends transcripts, the original file name, and your name, plus the summary format or your instructions if edited in Pro. For translation, only the spoken text from transcripts is sent, excluding timestamps and speakers.

For steps that use an OpenAI or Claude model for summarization or translation, Hearlog sends information to the OpenAI API or Anthropic API, respectively, using your API key after obtaining consent through 'Send to OpenAI' or 'Send to Claude'. Hearlog communicates only over HTTPS. For summarization, Hearlog sends transcripts, the original file name, and your name, plus the summary format or your instructions if edited in Pro. For translation, Hearlog sends only the spoken text from transcripts, excluding timestamps and speakers. Hearlog never sends recording files to either provider at any step. Content received through the API is not used for training by either provider. Hearlog does not request response storage from OpenAI. Content is retained by OpenAI for up to 30 days for abuse monitoring and deleted by Anthropic within 30 days.

For each transcription, summarization, or translation step that uses a model from a model server (Ollama or LM Studio), Hearlog sends the content to the server address you entered. For transcription, recorded audio is sent in segments separated by speaker, each lasting 20 seconds or less. For summarization, Hearlog sends transcripts, the original file name, and your name, plus the summary format or your instructions if edited in Pro. For translation, only the spoken text from transcripts is sent, excluding timestamps and speakers. If you entered a key, it is also sent. Hearlog does not obtain separate consent to send data to a model server. If the address uses http, data is sent without encryption. The speaker identification model used for transcription through a server is downloaded automatically from Hugging Face and may remain on the device until you uninstall the app.

Hearlog processes any transcription, summarization, or translation step for which you select a local model on this device. Recordings and transcripts for that step are not sent outside the device. Hearlog sends information to models selected for other steps and to save locations as described in this document.

Local models are downloaded from Hugging Face (huggingface.co). The request contains only the address of the file, and your device's IP address is shared.

Hearlog sends summaries and transcripts as Markdown files to each connected save location (a device folder or Google Drive). Transcript files also include translations and bookmark notes. The front matter at the top of each file includes the original file name, participants, keywords, and other details. File names include the recording time and summary title, and folder names include the group name. Recording files are also sent to save locations where 'Also send original audio' is enabled. If a cloud folder provided by another app is selected as the device folder, the files are uploaded by that app. Hearlog connects to Google Drive through Google account sign-in using Google Play services and receives permission to access only files created by Hearlog.

WebDAV servers, SFTP servers, S3-compatible storage, and OneDrive can also be connected as save locations. The username and password you enter are sent to the WebDAV server to connect. The username and password or private key you enter are used to sign in to the SFTP server. If the server key differs from the key used for the first connection, no files are sent until you approve the new key. A signature created with the access key ID and secret key you enter is sent to connect to S3-compatible storage; the secret key itself is not sent. OneDrive is connected by signing in to your Microsoft account, and access is granted only to the Hearlog app folder. Passwords and files are sent without encryption to WebDAV servers with 'Secure connection (https)' turned off. Files are sent without encryption to S3-compatible storage that uses an http endpoint.

If the app language is Korean or Japanese, Hearlog requests exchange rates from Frankfurter (api.frankfurter.dev) once a day to convert the estimated cost. The request contains only a currency code, and your device's IP address is sent with it.

To keep the model lists and estimated costs at current prices, Hearlog downloads a model price table from hearlog.umicorp.kr once a day. The request contains no recordings, usage history or device information, and your device's IP address is sent to Cloudflare, which hosts the site.

Free keys and keys from projects with billing enabled

Content sent with a free key and the responses may be used by Google to improve its products and may be reviewed by people. Google's terms advise against sending personal information with a free key. A key from a project with billing enabled is recommended because recordings contain conversations and names. If you enter both a free key and a paid key, the free key is used first by default. If you turn off 'Use free key' on the 'Gemini API key' screen, only the paid key is used for processing.

Content sent with a key from a project with billing enabled isn't used to improve products.

With either kind of key, Google keeps requests and responses for 55 days to prevent abuse, and authorized staff may review them.

Retention and deletion

Recordings in the app and downloaded local models are kept on this device until you delete them. Your usage records and the call recording folder's file list are kept until you uninstall the app.

Hearlog requests deletion of recording files uploaded to Gemini when processing of the recording finishes or is stopped. If there is no network connection, deletion is requested once a connection is available. Files that could not be deleted are deleted by Google 48 hours after upload.

Hearlog does not back up data to the cloud. When you move to a new phone through device-to-device transfer, recording files, API keys, sign-in details for save locations, and local models are not transferred. Other information, including transcripts, summaries, translations, your vocabulary, and settings, may be transferred. Device-to-device transfer may not be available, depending on the manufacturer and Android version.

All information stored in the app is deleted when you uninstall it. Files sent to save locations remain. Access granted to Google Drive also remains and can be revoked in your Google account settings.

Deletion and changes

Recordings can be deleted from the app along with their transcripts and summaries. Files saved to save locations are then moved to each save location's Trash folder or deleted immediately, depending on the settings. A local model can be deleted from the 'Local models' screen. Your vocabulary and API keys can be deleted in settings, and your name can be changed in settings. Your usage records and the call recording folder's file list cannot be deleted separately. When a Google Drive save location is deleted, Hearlog requests revocation of the access permission granted for Google Drive.

When you delete a WebDAV, SFTP, S3, or OneDrive save location, Hearlog deletes that location's sign-in details listed under 'API keys and sign-in details' below from this device. Access granted to OneDrive remains after you delete the save location or uninstall the app, so it must be revoked manually in your Microsoft account's app permissions.

API keys and sign-in details

API keys are encrypted with a key from Android Keystore and stored on this device. Hearlog does not store Google Drive access tokens and obtains them from Google Play services each time it sends files.

WebDAV and SFTP passwords, SFTP private keys and passphrases, S3 secret keys, and OneDrive sign-in tokens are also encrypted in the same way as API keys and stored on this device.

Withdrawing consent

Hearlog does not send recordings or transcripts to Gemini when you turn off 'Allow sending data to Gemini' on the 'Gemini API key' screen. Transcription and summarization using Gemini models wait until you turn consent back on. Translation using a Gemini model does not proceed if its provider differs from the summarization model's provider and consent is off when processing starts; otherwise, it waits. Hearlog may still send key verification requests and requests to delete uploaded recording files while consent is off. Processing for steps using a local model and sending files to save locations continue.

If you turn off 'Send to OpenAI' or 'Send to Claude' on the 'OpenAI' or 'Anthropic Claude' screen under 'Settings > AI connections', Hearlog stops sending information to that provider. Summarization using that provider's models is put on hold until you turn consent back on. Translation using that provider's models is not performed if the translation model's provider differs from the summarization model's provider and consent is off when processing starts. Otherwise, translation is put on hold until you turn consent back on.

Policy changes

Hearlog notifies you in the app when this privacy policy changes. Hearlog does not send information to AI services that require consent to send data, including Gemini, until you agree to the updated privacy policy.