Course Setup
Set metadata + modules. (Prototype stores state in localStorage.)
Identity
Modules
Script & Course Builder
Paste/write script, then use the tool panels to insert markers.
Click a chip on the right to insert a marker at your cursor.
Avatar Actions (click to insert)
Inspector
Active module
—
Chars
0
Modules
0
[AVATAR: gesture=Wave intensity=50]
[MEDIA_START: file=office.jpg placement=top-right duration=0]
[MEDIA_END]
[PAUSE: ms=400]
[PACE: slow]
[MEDIA_START: file=office.jpg placement=top-right duration=0]
[MEDIA_END]
[PAUSE: ms=400]
[PACE: slow]
Media Library
Mock library. Click “Insert to Script” to drop START/END markers.
Details
Select an asset.
Avatar
Mock avatar chooser + settings.
Avatar Gallery
Avatar Settings
Preview placeholder (real 3D preview comes later).
3D
Record Training
Prototype: simulates live transcript and converts it to script markers.
NEW (placeholder): “Record” can also mean Compile & Cache pre-generated LLM output + TTS audio + timing, so playback uses zero tokens. Q&A can still be live-token.
Live Session Capture (existing prototype)
Sessions (existing prototype)
Compile & Cache (token-saver placeholders)
This simulates creating a “compiled build” with:
manifest.json + per-line audio files + cue timings.
In production, these would be written to the published course folder (static-first).
Player behavior (future):
- Playback mode: read manifest + play audio files sequentially (no LLM calls).
- Interrupt: learner raises hand → live Q&A (tokens) → resume at same cue index.
No compiled builds yet.
Compiled Builds (placeholder list)
AI Assist
Prototype: generates mock script. (Real AI will be server-side.)
Preview
Mock learner view + validation warnings.
Learner View (Mock)
Later: this becomes embedded shared player using a real course.json URL.
Publish
Build course.json preview + simulated output plan.
Click “Simulate Publish”…
course.json Preview
Click “Build course.json”…
Settings / API
Prototype only. For SaaS/Enterprise, API keys should be stored server-side and never shipped to the browser.
Note: Browser TTS quality and voices vary by browser/OS and are not consistent.
We strongly recommend using Google, ElevenLabs, Azure, or Polly for consistent results.
If a selected provider is missing required keys, the builder will fall back to Browser TTS.
Browser TTS: Free fallback. We attempt to pick a good voice automatically, but it will vary by device.
Enterprise later: disable AI, enforce asset whitelists, SSO, on-prem runtime.