One credit system.
Every translation provider.
Skip the six API accounts. Comvi credits work with Azure, AWS, Google, DeepL, OpenAI and Anthropic — one balance, mix and match.
Five things to know.
- 01 Universal credits — one balance, every provider
- 02 Traditional MT — charged per character (1–2.5 credits/char)
- 03 AI translation — charged per token (~26–174 credits/translation)
- 04 Mix & match — different providers for different jobs
- 05 Monthly reset — credits refresh each billing cycle
Credits are universal.
Instead of managing separate accounts with Google, DeepL or OpenAI, you get a single credit balance that works across every provider Comvi supports.
What’s included
- Free plan — 20,000 credits / mo
- Early Adopter — 200,000 credits / mo
- Add-on — +100,000 credits for $1.50/mo
How it cycles
Credits reset at the start of each billing cycle. Unused credits don’t roll over — use them or lose them.
Per-character engines.
Fast, reliable engines — charged per source character. Good for high volume and short UI strings.
"Welcome to our application" — 27 characters
- Azure → 27 × 1 = 27 credits
- Google → 27 × 2 = 54 credits
- DeepL → 27 × 2.5 = 68 credits
Per-token models.
Higher quality translations from frontier LLMs — charged per input + output token.
Tokens are how AI models process text. Token count varies by language — English uses fewer tokens than Chinese or Arabic. AI translation has both input tokens (your text + system prompt) and output tokens (the translation).
* “per translation” assumes ~300 input tokens (text + system prompt) and ~8 output tokens (translated string).
With 20K free credits, that’s ~770 strings on GPT-5.4 mini.
Estimate your usage.
Plug in volume + average string length to compare credits across providers.
* Estimates are approximate. Actual usage depends on text complexity, language pair, and model tokenization. AI providers charge based on actual token counts returned by the API.
Pick the right tool.
Both belong in your toolkit. Here’s a quick decision guide.
Use traditional MT for
- ›Short UI strings — buttons, labels, tooltips
- ›High volume, simple content
- ›When speed matters most
- ›Budget-conscious translation
Use AI for
- ›Marketing copy & descriptions
- ›Nuanced or creative text
- ›Quality-critical translations
- ›When traditional MT sounds robotic
Credits, answered.
Translation requests are blocked until your credits reset (next billing cycle) or you buy a credits add-on. Existing translations stay accessible — you just can’t create new ones.
Budget AI models like GPT-5 mini have very low per-token rates. For short strings, the token count is minimal — making AI cheaper than per-character pricing. But AI has a fixed overhead (the system prompt), so for very short strings or bulk operations, traditional MT can be more efficient.
No — credits reset each billing cycle. This keeps pricing simple and predictable. If you consistently need more, add the credits add-on ($1.50/mo per 100K credits).
Input tokens include your source text plus the system prompt (translation instructions). Typically ~250–350 tokens for a standard translation request.
No hidden fees. The credit rates shown include our margin — what you see is what you pay. Your subscription covers platform access; credits cover translation API costs.
Start free with 20,000 credits.
No card required.
Upgrade to Early Adopter for 200,000 credits/month and lock the price forever.