Voice assistants are usually a separate product hiding inside the product: a different model, a shorter memory, and fewer tools. The moment you need real work done, you fall back to typing.
We built Aii Cloud Voice around the opposite idea. Voice is another way into the same Aii conversation.
Tap the microphone in a cloud chat and start talking. Your words appear as live captions. Aii answers aloud with synchronized text. When the conversation needs web search, reasoning, or a deliverable, the same agent loop does the work behind the voice.
Interrupt like a real conversation
You do not need to wait for a long answer to finish. Start speaking while Aii is talking and playback stops; your new turn takes over.
That matters for practical use. You can correct a wrong assumption halfway through, narrow a search before it grows, or say “skip the explanation and build it” without hunting for a stop button.
The transcript records the conversation that actually happened. It does not pretend you heard the part you interrupted.
Voice can produce visual answers
Some answers should not be spoken line by line. Ask for a chart, a document, or a small app and Voice can move into agent work, tell you it is working, and place the result in the conversation as an artifact card.
Open the card without ending the call. On supported devices, the voice controls stay available beside the workspace so you can inspect the result and keep talking about it:
“Make the labels larger.”
“Use the last twelve months instead.”
“Now turn this into something I can share.”
The result remains part of the same thread rather than becoming a disconnected voice-only answer.
Hang up without losing the work
When the call ends, the conversation is already a normal Aii Cloud chat. Continue by typing, reopen an artifact, share the thread, or delete it like any other conversation.
Aii Cloud stores the text transcript and artifact references. It does not store replayable audio. Speech frames are processed for the live session, while the durable record is the conversation itself.
Before a call starts, Aii shows the approximate credit rate and available balance. During the call, a live counter keeps the cost visible. If a connection fails, the partial transcript is preserved and the session ends honestly instead of leaving an endlessly spinning orb.
Built for the cloud lane
Voice lives in ordinary Aii Cloud chats. It is intentionally separate from a machine-controlled conversation, where microphone audio and local tool approvals would create a different trust boundary.
That separation keeps the promise simple:
- In Voice, you are talking to the cloud agent.
- In a connected-machine conversation, work and approvals stay tied to that machine.
Try it
Open Aii Cloud, start a cloud chat, and tap the microphone. Ask a question first. Then ask it to search or make something you can open.
The important moment is not hearing an answer. It is hanging up, seeing the complete thread still there, and continuing the work in whichever medium fits next.
Aii Cloud Voice is available in cloud chats. Audio is not retained; normal transcript, artifact, credit, and deletion controls still apply.