Songs written from journal entries

Privacy Policy

Last updated November 21, 2026

Feel It Back is a journaling and music app operated by Kitlinski Labs LLC (“Feel It Back”, “we”, “us”). This policy explains what personal data and content we collect on feelitback.com and in the Feel It Back apps, how we collect it, all uses of that data, when your content is shared with third-party AI or music services after your permission, and how to delete it.

The short version: your journal is private. Writing and saving an entry never send data to AI. Before using any AI feature, the app asks your explicit permission, disclosing the specific content sent and the third-party AI services that can receive it. Every third party receiving user data provides the same or equal protection described in this policy. We do not sell your data, show ads, or train AI models on your writing.

What we collect

  • Account information (collected via sign-in). When you sign in with WorkOS or Sign in with Apple, we collect the account details provided for authentication: email address, verification status, name, and profile picture when available. We also assign your internal account IDs and a display name you can edit.
  • Your content (provided directly by you). Journal entries, marks, AI reflections, photos, videos and recordings attached to entries, the feelings you write about them and their AI readings, song and lyric chat messages, lyric-memory notes, lyric versions, sound and style settings, voice personas, generated audio, cover images, playlists, and anything you add to a public profile.
  • Permissions, moderation, and preferences. Records of explicit AI data-sharing permissions you grant or withdraw, explicit lyrics and lyric memory settings, moderation reports, and user block selections.
  • Billing and usage (collected via Apple / RevenueCat and Stripe). Your plan, subscription status, and song allowance and usage. In the iOS app, Apple processes subscription payments as in-app purchases and RevenueCat verifies subscription entitlements using only an opaque account identifier (never your email, name, or journal content). On the web and in the Android app, Stripe processes payments; we receive subscription and invoice status, not your full card number.
  • Service and crash diagnostics. Generation status, error categories, moderation reports, and scrubbed crash reports via Sentry (release builds strip PII, journal text, lyrics, and prompts).
  • Product analytics and feature flags. The apps use PostHog with a random browser or device identifier to evaluate feature flags and collect an allow-listed set of scalar events without journal text, names, emails, or account IDs.

When your writing is sent to AI services

Before sharing data with any third-party AI service, we ask for your explicit permission on a dedicated disclosure screen that states what data can be sent and which AI services may receive it, and is versioned: if either list changes we ask permission again. You can withdraw permission and turn AI off at any time in Settings. That stops all future AI requests, cancels pending reflections, and clears song assistant threads. Requests do not include your name, email, or account ID.

Text requests run through the Convex AI Gateway or OpenRouter. OpenRouter requests use zero-retention and no-training controls. Gateway-routed requests do not currently send those same headers. The AI model companies that can receive the required content are Anthropic (Claude models), OpenAI (draft checks, backup lyric writing, and memory embeddings if enabled), and Google (Gemini models for themes, reflections, and media readings). Each feature sends only what it needs:

  • Generate lyrics and one-shot songs. These send your complete entry, including marked text (marks guide the draft; they do not hide anything), plus your theme, style settings, enabled lyric-memory notes, and the feelings and readings attached to the entry’s photos and videos (unless you turn that off). One-shot creation and style suggestions also send the entry so the model can recommend sound and lyric styles.
  • Lyric memory. After a first draft, the entry is sent again so the model can save up to three short notes about you, such as a recurring person or place. These notes help future songs. You can review them, delete them, or turn lyric memory off in Settings.
  • Automatic follow-on work. Once AI is on, making or changing lyrics also sends the lyrics (and the sound summary) in the background to summarize what changed, name the song and tag it.
  • Go deeper reflections. Each request sends the saved title and entry and up to three earlier reflections on that same entry. Other entries and lyric memory are not sent.
  • Song chat and the song studio. These send the song’s title and approved lyrics, your question, recent messages in that conversation, and a bounded set of your current and previous journal entries as context (up to 20 entries and about 18,000 characters for song chat). The journal context is not stored in the chat. The studio also sends its drafts and the titles, tags and lyrics of songs you reference.
  • Lyric edits and annotations. These send the song’s current lyrics, the lines you pick, your request and annotations, and the sound settings, without the journal entry. We also send the before and after lyrics to create short change summaries.
  • Style suggestions and artist styles. Suggesting settings sends the entry and your current settings. Typed sound directions and the artist and track names you choose, plus your current style settings, also go to the AI company, or to the music provider’s style-boost feature when it is on.

AI reflections and suggestions are stored separately from your writing. Writing and saving work without the disclosure. Declining it, or turning AI off later, means no request is sent to any of these companies; we never turn it back on for you.

When content is sent to music services

When you create a song, the music provider receives the approved lyrics, sound direction, and song title. It does not receive your journal entry or account ID. We currently use Kie.ai and may use MiniMax, fal, or Mureka instead. Cover images come from the music provider with each take.

Lyrics and sound direction come from your journal, so they may still contain names, events, or other personal details. You can review and edit them first. Create song in one shot skips that review and sends the generated lyrics and direction straight to the music provider.

If you create a voice persona, its name and description go to the music provider. If you enroll your own voice, which needs its own permission, a link to your recording goes to the provider as well. We cannot delete personas or voices from the music provider’s systems.

Voice input

Voice input is optional. The app asks for microphone and speech permission only when you choose to record. The native Dictate control uses on-device speech recognition and does not upload or save audio. If you choose Audio input, the recording is sent from our servers to the transcription provider configured for this app (OpenRouter by default, using Microsoft’s transcription model, or OpenAI) only under the AI disclosure you agreed to, the same one permission that covers photo readings and lyrics. The app shows that disclosure before the first transcript if you have not already agreed to it. The recording is stored privately on our servers; when you add its words, it is saved with your entry. In a web browser, the browser’s own speech service may send audio to its provider (for example, Google or Apple) for transcription. We do not receive or store that audio and cannot control how long the provider keeps it. Web voice input uses its own permission prompt for each recording and does not upload audio to our transcription provider.

Transcripts stay in the editor until you add them to your entry. Voice input never requests a reflection or a song on its own.

Photos and videos

Photos and videos you attach to an entry, and the feeling you write about them, are stored privately with that entry. If you tap Describe, each photo or video you chose and your feeling are sent from our servers to the AI provider configured for this app, which writes a reading you can edit. That provider is OpenRouter, using Google Gemini models by default, or OpenAI; these requests do not go through the Convex AI Gateway. The app shows the AI disclosure before the first reading. It is the one permission that also covers the lyrics features, so agreeing to it also lets the lyrics AI read your entry when you make a song. Save without a reading sends nothing to AI.

Earlier versions of the app describe a photo or video as a passage for your writing, with any context you type. Our servers now require the same AI disclosure permission for these requests too; older builds without a permission prompt may need an update before they can proceed. The media is stored privately on our servers and, when you add the words, saved with your entry.

A reading stays with its photo or video. Attached media, feelings and readings are never sent to music providers. When you make a song, the AI that writes lyrics receives your feelings and readings as notes kept apart from your writing, unless you turn off “Use photo readings in lyrics” in Settings; the photos and videos themselves are not sent. A reading you choose to add to your writing becomes part of your entry and is treated like the rest of it, so the AI features above receive it with your entry. Attachments, feelings and readings are not included in anything you share. Deleting an attachment, its entry, all your content or your account deletes them.

Optional connections

These apply only if the feature is offered to you and you turn it on.

  • Spotify. Imports are paused. If you previously connected Spotify, your saved top artists, tracks and encrypted connection tokens can be removed in Settings.
  • SoundCloud. Uploading a song sends its audio, title, and sharing setting to your SoundCloud account. Lyrics and journal text are not sent. Tracks on SoundCloud stay there until you delete them on SoundCloud.
  • Text messaging. If you link a phone number, messages you send us through Sendblue (iMessage/SMS) are stored for up to 24 hours before their content is cleared. If you allow voice-note transcription, voice notes are sent to OpenAI to be transcribed.

What other people can see

Nothing you write is public unless you publish or share it. Your journal text, feelings, AI readings and conversations, private notes, and prompts are never included in anything you share. Photos from an entry appear only if you include them (see below).

  • Song links are unlisted by default and show only the title and audio. You choose whether to add lyrics, artwork, or a note, and whether to list the song on your public profile. A link is a snapshot: later edits do not change it. Public pages ask search engines not to index them.
  • Photos on a song page. Where this option is offered, you can add photos from the song’s journal entry to its song page. Each photo stays private unless you include it and confirm. We publish a copy with location, camera, and other embedded details removed, along with any caption you type. Photos never appear in link previews. Removing the photo from your entry, or deleting the entry, removes it from the song page.
  • Public profiles show the handle, name, bio, picture, links, and public songs you choose.
  • Playlists are private by default. Shared playlists can be played by anyone with the link. Public playlists also appear in public browsing. Listeners see the title, description, track titles, take labels, and audio. Playlists are live, so your changes show up for listeners.

Unpublishing, making something private, or deleting it stops new access. It cannot recall copies someone has already saved. Anyone can report a public or shared song, a public profile, or a public or shared playlist. Reports and moderation records are deleted after 90 days.

How we use data

We use your data to:

  • run the features you use
  • keep your content in sync across devices
  • track your plan and allowance
  • fix failures and provide support
  • prevent abuse and meet legal obligations

Access to user content is limited to operating the service. Admin tools show account and billing details but not journal entries or lyrics. We only look at private content when it is needed to investigate a problem you report, protect the service, or comply with the law.

We do not use your writing, lyrics, or songs to train AI models. If that ever changes, it will be a separate opt-in. We do not sell or rent personal information, and we do not share it for targeted advertising.

Service providers

We rely on these third-party services and processors. Each one receives only what its part of the service needs:

  • AI model and routing providers (only after your explicit permission): Anthropic (Claude), OpenAI, Google (Gemini models, including theme, reflection, and media reading), Convex AI Gateway, and OpenRouter
  • WorkOS: authentication and account identity
  • Apple and RevenueCat: iOS in-app purchase processing and subscription entitlement verification (RevenueCat receives only your opaque account ID and Apple transaction status, never your email, name, or journal)
  • Stripe: payments on the web and in the Android app
  • Vercel: website hosting and social-video rendering
  • PostHog: feature flags and privacy-safe product analytics
  • Sentry: crash and performance diagnostics (scrubbed of journal text, lyrics, prompts, and PII in release builds)
  • Music generation (only approved lyrics and sound direction): Kie.ai, MiniMax, fal, or Mureka
  • Photo, video and audio AI, only when you ask for a reading or transcription: OpenRouter (Google Gemini models for photos and videos and Microsoft’s model for recordings, by default) or OpenAI
  • Only if you connect or use those optional features: SoundCloud, Sendblue

Same or equal protection by third parties

Every third-party AI service and service provider with which we share personal data or user content is contractually and technically required to provide the same or equal protection of user data as stated in this Privacy Policy and required by Apple App Store Guidelines. Third-party AI providers receive requests solely to perform the feature you requested under zero-retention and no-training controls where available, may not use your journal entries, lyrics, recordings, or media to train shared foundation models, and may not sell or repurpose your data. Providers may process data in the United States and other countries subject to those equal-protection safeguards.

On your device

The iOS and Android apps keep your journal and recent songs available offline. Text and sign-in tokens go in the device’s secure storage. Downloaded audio and artwork go in the app’s storage folder.

Unsaved drafts are backed up locally so you can recover them after an interruption. In the apps, the backup uses secure storage. On the web, it uses the browser’s local storage. A backup is removed when the text is saved or discarded. Signing out or deleting your account clears the app’s offline copy on that device.

Retention and deletion

We keep your content until you delete it or delete your account. You can delete individual entries, songs, and playlists at any time.

Delete all content in Settings removes your entries, reflections, marks, memory notes, songs, lyrics, audio, playlists, and AI permissions. It also clears your public profile. Your sign-in, billing records, and song usage stay, and deleting content does not refund songs already used.

Delete account in Settings permanently removes your account and everything above: public songs, profile, connections, billing records, and your WorkOS sign-in. This cannot be undone, and there is no grace period.

Before deleting your account, cancel any active subscription: in your App Store subscription settings for purchases made in the iOS app, or in billing settings on the web or in the Android app. Deleting the app does not cancel it. After deletion, we keep only a minimal record that the deletion happened, admin audit logs, and moderation records for up to 90 days. Some things are outside our control and are not deleted:

  • Apple’s and Stripe’s own payment records
  • tracks you uploaded to SoundCloud
  • voice personas held by the music provider
  • data already sent to AI providers

You can download lyrics as text and songs as MP3 or WAV files. Keep your own copy of anything you want to keep outside Feel It Back.

Security

Data is encrypted in transit. Sign-in is handled by WorkOS, and access to your content is checked on the server for every request. No service is perfectly secure, so avoid writing passwords or financial details in your journal.

Your rights

Depending on where you live, you may have the right to access, correct, delete, or receive a copy of your personal information. You may also have the right to object to or restrict how we use it. Most of this can be done in the app. For anything else, email us. We will not treat you differently for using these rights. If you are in the EU or UK, you can also complain to your local data protection authority.

Children

Feel It Back is not directed to children under 13, and they may not use it. If you believe a child has given us personal information, contact us and we will delete it.

Changes

We will update this page when our practices change and revise the date above. If a change affects how your writing is sent to AI or music providers, we will ask for your permission in the app before it applies.

Contact

For privacy questions or requests, email kyle@kkit.dev. Please describe the issue first rather than including journal entries, lyrics, or payment details. See also our Terms of Service.

Feel It Back is not therapy, crisis care, or medical advice. If you may be in immediate danger, contact local emergency services.