We build software that feels human.
San Combisco is a digital company making realtime AI products — characters that see, speak and answer in milliseconds. We take them from the first idea to production, and we keep them running.
A few of the things we've shipped.
A sample, not a catalogue — the ones we can show. None of these are prototypes or case studies: every screenshot is the live product, running today.
What we are unusually good at.
Four things we have learned the expensive way, and now do on purpose.
Realtime, not nearly realtime
Ten milliseconds is the difference between a character and a chatbot. Everything we build is engineered backwards from the moment a person expects an answer.
Models where they should not fit
We take generative video and audio and make it run at scale in the cloud — or entirely on the phone in someone's pocket, with nothing leaving the device.
From nothing to live
Everything we ship begins as an empty repository. We do the whole arc: the idea, the thing people touch, and the machinery underneath it.
We stay and run it
Launching is the easy half. We own what we ship through the scale, the cost curve, and the second year — where most software quietly falls over.
- Character response latency
- 0ms
Character response latency
- Concurrent conversations
- ∞
Concurrent conversations
- Data residency, GDPR by default
- EU
Data residency, GDPR by default
Latency, concurrency and residency as published for Equos characters.
Tell us what you want to build.
A realtime AI product, a model that has to run somewhere it shouldn’t fit, or an idea still on a napkin. We read everything.
Or write to contact@san-combisco.com