Private voice journaling, on your device or your own AI key
Journaling works best when the space feels contained - when you're not editing yourself for an audience that might be listening. Keep recording, transcript, and search on the device you're holding, so you can think out loud without managing your own surveillance.
On your device
When to keep journaling on-device
This is the first-choice setup for raw reflection: founder brain dumps, end-of-day reviews, the thoughts that aren't ready for a shared workspace and might never be. Recording, transcription, storage, and search all happen in the browser workspace on the current device. Nothing leaves unless you choose an export, backup, BYOK summary, or switch to a cloud provider.
For most journaling, that's all you need. You speak, the transcript lands locally, you search it later when you want to reconnect a thread of thinking. No background uploads, no shared workspace, no "your data may be used to improve our service."
How it works on your device
- Open jotty.pro on the device you journal on, with transcription on-device.
- Record the thought, reflection, or end-of-day review.
- Keep the transcript local while it's still raw - you don't owe anyone a clean version.
- Search local entries later when you want to reconnect ideas across weeks.
- Move to a cloud provider only if you intentionally want cloud transcripts and have reviewed the terms that govern them.
With your own cloud key
Already paying for OpenAI, Google, or another cloud provider for personal work? Bring that relationship to your journal: better summaries, deeper pattern detection, your key, transcripts still local.
When a cloud provider is the right call
On-device is the right default if you don't trust any provider with journal audio. But if you already trust a specific provider for personal work, routing your journal through them keeps that trust and gives you the richer self-reflection bigger cloud models can do.
In practice that means longer-form summaries, pattern detection across entries over weeks or months, and prompts that actually engage with what you've been writing. On-device models can't quite match that yet, and pretending otherwise would be marketing copy.
The boundary: we route your audio to the provider you pick, under your account. We don't see the audio, the transcript, or your key past the current session. Transcripts live in local DuckDB on your device, same as the on-device path, same as every other configuration.
How it works with your key
- Pick the provider you already trust: OpenAI, Groq, Deepgram, HuggingFace, Ollama, X.AI, Google, Mistral, or DeepSeek.
- Paste your personal API key in onboarding. We route to your account, not a shared pool.
- Record the journal entry the way you normally would.
- We send the audio to your provider for transcription, and optionally for a summary or reflective prompt.
- The transcript and any AI output write to local DuckDB on your device. Nothing lands on our servers.
How it compares
| Feature | Generic transcription apps | On-device transcription | jotty.pro with your personal cloud key |
|---|---|---|---|
| Who sees the journal text | The app vendor and any integrated partners | Only your device | Only your chosen provider, under your account |
| Who holds the key | Vendor manages credentials | No external key needed | You hold and supply your own API key |
| Model quality vs privacy | Vendor controls both | On-device quality; strongest posture | Cloud model quality; provider trust is your call |
| Where the journal lives long-term | Vendor account or cloud storage | Local DuckDB on the device | Local DuckDB on the device; not on our servers |
Honest answers
Is on-device enough for every private journal?
It's a strong default for device-scoped journaling, but you still control device security, backups, and any optional export or cloud-provider step. If your device walks out of the room with someone else, the strongest data-path posture in the world doesn't help.
What if I want my journal on another device later?
Treat that as a cloud-provider or export decision - review the applicable terms and decide whether cloud transcripts fit the sensitivity of the journal. Don't enable it first and ask questions later.
Can I use AI prompts on a journal entry?
Yes, when you pick a BYOK provider. The selected text you send is governed by that provider's terms, same as any BYOK call. Keeping it on-device just means the default isn't "send everything."
Why send a personal journal to any provider?
Because you might already be doing it. If you use OpenAI or Google for personal work and have read the terms, routing your journal through them doesn't introduce a new vendor; it extends the relationship you already have. If you don't trust any cloud provider with audio of yourself thinking out loud, keep it on-device.
Can I rotate or revoke my key later?
Yes. The key is yours. Rotate or revoke it any time through your provider's dashboard. We don't store or cache it beyond the current session. Once you revoke, new audio stops routing until you supply a fresh key.
Will the provider use my entries to train models?
We can't answer that for you. It depends entirely on the provider's data-usage policy and which tier you're on. Check their terms directly, especially the sections on API data usage and training opt-outs. Some providers offer zero-data-retention tiers or explicit opt-outs for API traffic. The provider is the only authoritative source on this; we don't have visibility into how they handle data once it leaves our routing layer.