Plans & limits
What you get on each plan
Swiftin has three plans: Free, Pro, and Team. The main differences: how many AI tokens you get per month and which features are unlocked.
What's a token?
A token is the unit AI providers count text in. Roughly: 1 token ≈ 4 characters in English, 1–2 characters in Chinese / Japanese / Korean. Numbers, code and rare words also take more tokens. Swiftin passes the provider's reported usage straight through to your quota — what they charge us is what gets debited from your pool.
How one AI translation is billed
Each request to a backend AI model is one billing event. The amount debited depends on the model's per-token input and output multipliers — every wired model (Gemini 3.1 Flash-Lite, Llama 3.3 70B, GPT-5 nano, Claude Haiku 4.5, DeepSeek V4 Flash, Grok 4.3) has its own pair, calibrated to that provider's real cost. Full per-model rate table at /docs/translation-credits.
Prompt tokens (input)
Everything sent to the model on your behalf: the short system instruction ("You are a professional translation engine."), the user-prompt template that wraps your text, and your text itself packed as a JSON array. For Page Translate, the prompt also carries the page Title and — when Smart Context is on for Pro/Team — a short AI-generated page summary so the model has context.
Completion tokens (output)
The translated text the model returns, wrapped in the same JSON array shape. Sent back token-by-token. Heavier languages (Cyrillic, CJK) usually produce more tokens than the input — that's a property of the tokenizer, not a Swiftin surcharge.
Debit per request = prompt_tokens × inputMultiplier + completion_tokens × outputMultiplier. Multipliers differ per model because providers charge input and output at different rates (output is typically 2-8× more expensive). Cheap models like DeepSeek burn your quota slowly; premium models like Grok burn it faster. Per-model table at /docs/translation-credits.
Other things that move your tokens
Two surfaces outside the translate request itself also draw from your pools.
Smart Context summary (AI pool, Pro/Team)
When you opt into Smart Context on a Pro/Team plan, the first user to translate a given URL pays a one-off ~300–500 AI tokens for the page-summary call. The summary is cached for 24 h, so anyone (you or someone else on Pro/Team) hitting the same URL in that window reuses it for free. Free users with the toggle on get a title-only context and pay nothing extra.
Text-to-speech (TTS pool, separate)
TTS has its own token pool, distinct from translation. Each playback debits prompt tokens (the text you asked us to speak) plus audio tokens (Gemini's accounting for the generated speech). The two are summed and pulled from your TTS quota.
When you're charged — and when you're not
The backend uses post-debit billing: nothing is taken until the provider has actually returned a useful result. The list below reflects what the translation pipeline does today.
Debits your pool
- •Successful AI translation through a backend engine (Gemini / Llama / OpenAI / Anthropic / DeepSeek / Grok) — debit equals prompt × inputMultiplier + completion × outputMultiplier per the model's rate (see /docs/translation-credits).
- •Successful Smart Context summary on a URL nobody else has summarised in the last 24 h — one-off ~300–500 AI tokens.
- •Successful TTS playback — prompt_tokens + audio_tokens from Gemini, debited from the TTS pool only.
Costs nothing
- •The provider call fails (auth, 429 rate-limit, 5xx, network drop) — telemetry records 0 debited.
- •Source and target turn out to be the same language (the model returned the input back) — caught after the call, before the debit.
- •The translation is already in your local cache — IndexedDB serves it, the backend isn't called at all.
- •You used Google or Bing — free engines never touch the AI-token pool.
- •Smart Context cache hit — another Pro/Team user paid for that URL's summary within the last 24 h.
- •BYOK Custom mode — the extension calls the provider directly with your key; Swiftin never sees the request and never bills.
| Feature | Free | Pro | Team |
|---|---|---|---|
| AI tokens / month | 50K | 5M | 10M / seat |
| TTS tokens / month | 1K | 50K | 50K / seat |
| Page / Selection / Input translation | ✓ | ✓ | ✓ |
| Chat translation (8 platforms) | — | ✓ | ✓ |
| Translation styles (Slang, Business) | — | ✓ | ✓ |
| AI Smart Context | — | ✓ | ✓ |
| Translation history | — | Opt-in, 90d | Opt-in, 180d |
| Cloud sync | — | ✓ | ✓ |
| Google + Bing fallback | Unlimited | Unlimited | Unlimited |
Monthly reset
Quotas reset on your billing anniversary (or 1st of the month for Free). Unused tokens don't roll over. Need more? See addon top-ups in the dashboard.
See plans and pricing