Privacy

Your journal is yours. Period.

We built Anima with one rule: your most personal data should stay personal.

Voice recordings never leave your phone. The AI sees only what it needs and forgets it immediately. You can delete everything at any time.

How your data flows

When you record a voice entry, here is exactly what happens:

Step 1
You speak
Step 2
On-device transcription
Step 3
Transcript sent to Groq AI
Step 4
Stats returned to your device
Data type Where it lives Stored by AI?
Voice recording (audio) On device only Never sent
Transcript text On device Processed, not stored
Stat scores and XP On device Never sent
Mandala and title On device Never sent

Voice recordings stay on your device

Your voice recording is processed into text right on your phone using on-device speech recognition. The raw audio file is never uploaded anywhere. Once the transcript is generated, the audio is available only to you, stored locally on your device.

We chose this approach because voice is the most intimate data you can share. It carries tone, emotion, and identity markers that go far beyond what words on a page reveal. That data belongs to you and only you.

What this means in practice

If someone gained access to Anima's servers, they would not find a single voice recording. Not yours, not anyone's. The audio physically does not exist outside your phone.

Transcript text is classified by Groq AI

Once your voice is transcribed to text, that transcript is sent to Groq AI for activity classification. Groq runs Meta's Llama large language models to identify the activities in your entry and map them to Anima's seven human stats.

We chose Groq for two reasons: speed and privacy. Groq processes requests in real time with extremely low latency, and their infrastructure is designed for stateless inference.

What Groq sees

Groq receives the text of your entry and a classification prompt. It returns structured data (activity names, stat categories, XP values). That is the entire interaction. No names, no account IDs, no device identifiers are included in the request.

Groq does not store your data

Groq's API is stateless. Your transcript text is processed in memory and discarded after the response is returned. There is no training on your data, no logging of inputs, and no retention beyond the processing window.

This is not a policy promise that could change later. It is an architectural constraint of how Groq's inference engine works. The data passes through and is gone.

  • No training on user data. Groq's Llama models are pre-trained. Your entries do not influence the model.
  • No input logging. Requests are not stored in any database or log file on Groq's infrastructure.
  • No cross-user data sharing. Each request is independent. There is no profile built from your entries.

Delete your data or go fully offline

You are in control. Anima gives you two ways to take ownership of your privacy:

Delete everything

From the app settings, you can delete individual journal entries, all entries at once, or your entire account. Deletion is immediate and permanent. Since your data lives on your device and Groq does not retain anything, there is nothing left anywhere.

Switch to offline mode

If you prefer that absolutely nothing leaves your phone, you can enable offline mode. In this mode, all classification happens on-device using a local model. No network requests are made at all. Your entries, transcripts, and stats stay entirely on your phone. Classification accuracy may differ slightly from the cloud model, but your data never touches the internet.

What we do and do not collect

We collect We do not collect
Email address (if you signed up for the waitlist or created an account) Voice recordings
Anonymous usage analytics (screens visited, feature usage) Journal transcript text
Crash reports Stat scores or XP data
Location data
Contacts or photos

Common questions

What if Groq changes their policy?

If Groq ever changed their data handling practices, we would migrate to another stateless inference provider before the change took effect. Our architecture is provider-agnostic. The classification prompt and response format work with any Llama-compatible API.

Do you sell data to advertisers?

No. Anima has no advertising. We do not sell, share, or monetise your personal data in any form. Our business model is the app itself, not your data.

Is my data encrypted?

Yes. Data on your device is protected by iOS device encryption. Data in transit to Groq is encrypted via TLS 1.3. At no point does your data travel unencrypted.

Can I export my data?

Yes. You can export all your journal entries, stats, and character data from the app settings at any time.

Disclaimer

Not professional advice

Anima provides personal reflections and insights based on your own journal entries. It is not a substitute for professional medical, mental health, financial, or legal advice. The AI-powered stat agents and insights are designed for self-reflection and personal growth, not diagnosis or treatment. If you are experiencing a medical or mental health emergency, please contact your local emergency services or a qualified healthcare professional immediately. Always consult appropriate professionals before making decisions about your health, finances, or legal matters based on any content within Anima.

Ready?

Your journal. Your data. Your character.

Download Anima and start building the most private journal you have ever kept.

Download Free