What Actually Matters on iPhone and iPad

Before comparing apps, separate the requirements that break most mobile workflows:

• Card compatibility. Character cards (PNG with embedded JSON, or plain JSON) must import without losing description, personality, scenario, first_mes, and alternate_greetings. Lorebook/World Info support is the usual differentiator.
• Backend freedom. You need to point the client at OpenAI-compatible endpoints, OpenRouter, or a local proxy — not be locked into one vendor’s models.
• Prompt control. System prompt editing, sampler settings, and the ability to keep a SillyTavern-style prompt structure matter more than UI polish.
• Persistence. Chat history and card libraries must survive app updates and device changes.

Anything that fails two or more of these is a demo, not a tool.

The Realistic Options

1. Self-hosted SillyTavern via Safari or a PWA

You can run SillyTavern on a home server, VPS, or a machine on your LAN and open it in Safari. On iPad it behaves acceptably; on iPhone the UI is cramped but usable in landscape.

• Pros: full feature parity, your existing cards and lorebooks work unchanged.
• Cons: requires a always-reachable host, HTTPS for some features, and no offline use. Backgrounding Safari can drop long generations.

This remains the reference implementation. Everything else is judged against it.

2. Native iOS frontends

A handful of native apps now parse SillyTavern-format cards directly. Quality varies widely; check for World Info support and whether the app exposes raw sampler parameters. Many “AI roleplay” App Store apps are closed ecosystems that strip card metadata and force their own models — avoid these if you maintain a card library.

3. MiniTavern (iOS, Android, Web, Chrome extension)

MiniTavern is built around the same card format and workflow SillyTavern users already know: import a card, attach lorebook entries, choose a backend, chat. The iOS and iPad apps are the relevant part here, but the same account and card library carry over to the web client and the Chrome extension, which is useful when you draft a character on desktop and continue on a phone.

• Native card import for PNG/JSON, including greeting variants.
• Bring-your-own API key against OpenAI-compatible endpoints.
• A card market for discovering and sharing characters without manually exporting files.

If your objection to hosted apps is lock-in, the test is simple: export your card and confirm nothing was stripped.

4. Local inference on-device

Possible on recent iPhones and iPads with small quantized models, but context length and quality still fall short of what card-driven roleplay needs. Treat it as an offline fallback, not a primary setup.

Comparison at a Glance

Option	Card fidelity	Backend choice	Offline	Setup effort
Self-hosted ST in Safari	Full	Full	No	High
Native third-party apps	Varies	Often limited	Partial	Low
MiniTavern	Full	OpenAI-compatible	No	Low
On-device LLM	Full	Local only	Yes	Medium

A Workflow That Holds Up

1. Keep a canonical card library in a synced folder or a market account — never only inside one app.
2. Store API keys per backend, not per character, so switching models doesn’t mean re-editing cards.
3. Test every new client with one demanding card before migrating anything: multiple greetings, a lorebook with keyword triggers, and a long first message.
4. Export chat logs periodically; JSONL exports are the most portable.

Featured Card: Captain Elara Voss, the Starfall Navigator

A good stress test for any iOS client is a card with layered lorebook entries and conditional greetings. Captain Elara Voss, the Starfall Navigator fits that profile: a void-faring ship commander whose scenario depends on consistent world details — star charts, crew hierarchy, jump-drive mechanics — that only hold together if the app actually loads World Info on keyword triggers.

Import her on iPhone or iPad and check three things: does the opening greeting render fully, do lorebook entries fire when you mention the ship or the drive, and does an alternate greeting load without duplicating history. If all three pass, the client is worth keeping.

Conclusion

For SillyTavern users on iOS, the choice comes down to how much of your existing workflow you’re willing to rebuild. Self-hosting keeps everything but costs convenience; native apps save time but often sacrifice card fidelity. MiniTavern sits in between — the same character cards and lorebook structure, native iPhone and iPad apps, a Chrome extension for desktop drafting, and a card market for finding new characters. If you want to test it against your own library, start with Captain Elara Voss in the MiniTavern card market and see whether the import survives intact.