ERAIS · unified world
Unified World
Start a session. Available experiences appear when the service connects.
Checking the session backend
Enter to send · Shift+Enter for a new line. Input options depend on the connected runtime.
Make something new
Your words. Spoken here.
Write a line, choose a route and hear a fresh recording.
Checking the shared ERAIS service…
What runs when you press Generate?
Both routes run through the shared ERAIS service. The Fracture route executes the two converted projections in one speech decoder block using the ERAIS native engine. Other speech components still come from LuxTTS and Vocos. The reference uses their original projections. This is an experimental conversion, with no claim that the whole speech model is native or that every voice is improved.
About this prototype & your privacy
The backend label identifies the connected service. Available controls reflect its enabled capabilities. The status below follows the same owner and its private learning connection. Confirm any offered 10-minute renewal to keep your session.
Explicit web lookups send only your current public topic to search engines. Links identify snippets, not verified answers.
Only inputs you enable or submit reach the runtime. This tab compresses camera frames and uploaded images to ≤512 pixels on their longest side and strips metadata. Changing scenes trigger frames; audible sound triggers short microphone clips, which may miss speech. Attached images and supported audio stay in this tab for follow-up questions until removed or the session ends.
Router learning’s private candidate is discarded at session end; live routing stays unchanged. Ending closes your session and stops capture. Hiding or leaving also stops camera and microphone. Browser read-aloud uses your device’s speech service, separately labelled from model-generated speech. Session data notice.
Research activity and service details
Behind your session
One service. Separate visitor and research state.
The same desktop service handles your conversation and speech. Visitor chat stays within its session and does not train the model. The saved connectome state shown here belongs to a separate private research episode.
Checking for a fresh status update…
- Requests in progress
- — Waiting for current capacity.
- Private observations processed
- — State saved after each accepted input
- Private learning connection
- — Native connectome vision and synaptic state
While you watch
- Waiting for the first live response.
The last 10 updates stay in this tab. Pausing this view leaves the service running.
What is connected, and what is still being built?
Conversation, camera comparisons, speech and the private learning episode use one service owner and a shared work limit. The learning connection accepts small visual observations and bounded simulation inputs. Native conversation is experimental. Camera input compares your view with 80 everyday object categories and offers uncertain hints. It does not locate, count or identify people. The current vision route uses every imported expert; sparse savings, general audio generation and game migrations remain in progress.
- Model requests
- —
- Native learning requests
- —
Your session inputs are not used for training. These counts show completed work, not improved intelligence or game skill. Explore the applications or read the evidence.