One API, 90+ models, one bill

Switching between model providers usually means a second account, a second key, a second invoice and a second SDK. GlianaAI is one endpoint in front of ninety-odd models.

What is behind it

• Image — generation and editing, including multi-reference workflows
• Video — text-to-video and image-to-video
• Music and speech — generation, text-to-speech, transcription
• Chat — an OpenAI-compatible endpoint, so existing clients work unchanged
• Utility tools — web scraping, screenshots, structured extraction, search, embeddings, moderation
• Indonesian identity and data — KTP reading, face comparison, region lookup, exchange rates

The catalog is an endpoint, not a document

Models, prices and input schemas are published live. Anything calling the API can fetch the current list and the current price at runtime, rather than reading a spec that drifted a month ago.

That matters for automated callers: a program can discover what is available and what it costs without a human updating a config file.

Pricing is per call

No subscription, no per-seat cost, no per-environment cost. You pay for the calls you make, and the price for each is published before you make it.

For anything with a variable cost — a longer video, a bigger output — we quote the ceiling up front rather than surprising you afterwards.

Failures do not cost you

If a model fails, you are refunded. Requests are validated before payment is taken, so a malformed call is rejected free rather than charged and then failed.

Built for programs, not just people

There is no signup. A hosted agent exists if you want a chat interface, and there are MCP servers if your tooling speaks that — but the API itself expects to be called by software, and is priced and documented accordingly.