Most AI builder tools treat export as a finishing step.
You generate one result for one screen, then manually adapt it for every channel.
For teams shipping real products, that model fails fast.
You need one source of truth that can project to many surfaces:
What "multi-surface" really means
Multi-surface output is not copy-paste between templates.
It means the runtime holds one canonical scene state, and each surface renders it with its own constraints.
Surface constraints include:
Why this matters for AI app builders
An AI app builder that only supports one surface is still a prototype tool.
Production teams need:
Surface strategy we use
At Dreams.fm, we classify surfaces into three groups.
1. Interactive product surfaces
For live editing and end-user product interaction.
2. Broadcast and stream surfaces
For walkthroughs, demos, and collaborative sessions.
3. Narrative presentation surfaces
For structured story flow with controlled pacing and emphasis.
All three use the same scene state and timeline history.
Projection without duplication
To avoid duplicated logic:
This lets each surface evolve independently while preserving one runtime model.
Fidelity modes
Different surfaces need different quality and latency tradeoffs.
We use fidelity tiers:
The scene does not change between tiers. Rendering behavior does.
Common mistakes
Mistake 1: embedding surface assumptions in core state
Fix: keep surface metadata separate from canonical scene structure.
Mistake 2: rewriting transforms per surface
Fix: transforms should modify runtime state once; projection should adapt output.
Mistake 3: skipping compatibility tests
Fix: add surface-level validation to catch layout and interaction regressions.
Where fmEngine fits
Internally, fmEngine handles projection routing and timeline-safe transforms. Externally, teams get one practical advantage: generate and publish to multiple channels without rebuilding the project every time.
SEO and discoverability
"Multi-surface" alone is not a high-volume term, so we pair this concept with stronger category terms:
That keeps content discoverable while still teaching advanced runtime concepts.
Closing
If your AI builder cannot project one source state to multiple surfaces, scaling from prototype to production becomes expensive.
A runtime-first, projection-aware architecture gives teams leverage and consistency.
That is core to what we are delivering in Dreams.fm private beta.
