Skip to main content

Choosing your models

The Models page in your web account is where you decide which AI does the two jobs at the heart of InkSpoke: turning your voice into text, and refining that text into clean, polished writing. You set an active model for each, and your choice is saved to your account.

Find it at Account → Models once you're signed in.

The two choices you're making​

Every dictation runs through two stages, and this page lets you choose the model for each one:

  • Voice → text — the speech-recognition (transcription) model that hears your words and writes them down.
  • Text refinement — the language model that cleans up filler, fixes grammar, and matches the tone of what you're writing.

Each stage has its own list of models, so you can mix and match — for example, a fast on-device transcriber paired with a cloud model for refinement.

What the page looks like​

Models are grouped by provider. Each model shows a short description, a tier badge (free or Pro), and whether it's currently available. One model in each category is marked Active.

┌────────────────────────────────────────────────────────────┐
│  Models                       You're on Free · manage →    │
├────────────────────────────────────────────────────────────┤
│  Voice → text                                              │
│  ┌────────────────────────┐  ┌────────────────────────┐    │
│  │ ○ OpenAI               │  │ ● Whisper (on-device)  │    │
│  │   Cloud transcription  │  │   Runs locally  [free] │    │
│  │   [free]   [Set active]│  │            ● Active     │   │
│  └────────────────────────┘  └────────────────────────┘    │
│                                                            │
│  Text refinement                                           │
│  ┌────────────────────────┐  ┌────────────────────────┐    │
│  │ ○ Anthropic     [Pro]  │  │ ● Platform model       │    │
│  │   Advanced refinement  │  │   Included  [free]     │    │
│  │   Upgrade to Pro →     │  │            [Set active]│    │
│  └────────────────────────┘  └────────────────────────┘    │
└────────────────────────────────────────────────────────────┘

The header shows which plan you're on ("You're on Free · manage →" or "Pro"), with a link through to your subscription.

Set your active models​

  1. Go to Account → Models.
  2. Under Voice → text, find the model you want and click Set active.
  3. Under Text refinement, do the same for the model you want to refine your writing.
  4. That's it — your selection is saved to your account through the API, so it applies to your InkSpoke account rather than living only in one browser.

The active model for each category is highlighted so you can always see, at a glance, what's doing the work.

What the labels and buttons mean​

Label / controlWhat it means
ActiveThis model is currently handling that stage (voice or text).
Set activeClick to make this the model used for that stage.
Free / Pro badgeThe plan tier a model belongs to.
Pro chip + Upgrade to ProShown to Free users on a Pro-only model; links to your subscription page.
AvailabilityModels that aren't currently available are shown but can't be set active.

Free vs. Pro models​

Every model carries a tier badge. On the Free plan you use InkSpoke's built-in, platform-hosted models. Pro unlocks the extra models — and the ability to bring your own provider keys and download models to run on-device.

If you're on Free and click a Pro-only model, you'll see a Pro chip and an Upgrade to Pro prompt instead of a "Set active" button. It links straight to your subscription page, where you can move up a plan.

tip

You don't need Pro to get great results. Free accounts run on InkSpoke's platform models, and the desktop app's default on-device speech model (Whisper Small) is free and fully offline. Upgrade when you want a specific provider, larger models, or your own keys.

Downloading models and adding your own keys — do it in the app​

Two things on this page are entry points, not actions you complete on the web:

  • On-device models show a Download model button.
  • Bring-your-own-key (BYOK) providers show an Add API key button.
Download and Add API key happen in the apps

On the web, these buttons are placeholders. To actually download an on-device model or add your own provider key, do it in the desktop or mobile app, where model files and credentials are stored securely on your device. See Models and providers for the desktop flow.

What the web page does do reliably is let you choose the active model for each stage — that part is fully live and saved to your account.

Power-user notes​

  • Providers you'll see. The catalog is grouped by provider, and InkSpoke recognizes names like OpenAI, Anthropic, Google, Deepgram, AssemblyAI, ElevenLabs, Groq, Mistral, and Azure. The exact models available to you depend on your plan and what's enabled on your account.
  • Mix on-device and cloud. Because the two stages are chosen independently, you can keep transcription on-device for privacy and use a cloud model for refinement, or any other combination that suits you.
  • Manage keys separately. Your personal API keys are managed on their own page — they're for accessing your InkSpoke account (and unlocking your encrypted synced data), not for BYOK provider credentials, which live in the app.
Routing rules are a preview

The Models page shows an advanced Routing rules table for automatically picking a model per situation. It's a design preview populated with example rules — it isn't wired up yet, so treat it as a look ahead rather than a working control.

Next steps​