Guides
Roblox StreamingEnabled: best settings and common bugs
How Roblox StreamingEnabled works, the recommended settings, model streaming modes, and fixing scripts and parts that break when streaming.
3 min read

Quick answer
What does StreamingEnabled do in Roblox?
It loads parts of the Workspace onto each player's device as they get near and unloads them as they move away, instead of sending the whole map at once. That means faster joins and less memory, especially on phones. It's on by default for new places, but scripts must expect parts that haven't loaded yet.
How it works
With StreamingEnabled on, the server sends each player only the parts of the Workspace near them, and the client drops distant parts when memory is tight. It's on by default for new places, and you set it on the Workspace in Studio. Scripts can't change it (streaming docs).
The key settings
StreamingMinRadius: Default: 64 studs; What it does: Content inside this radius loads first and never streams outStreamingTargetRadius: Default: 1,024 studs; What it does: The furthest distance content streams in toStreamOutBehavior: Default: LowMemory; What it does: Roblox recommends Opportunistic, which removes distant content more readily to save memoryStreamingIntegrityMode: Default: See the docs; What it does: Roblox recommends PauseOutsideLoadedArea, which pauses a player who reaches an unloaded area
Keep StreamingTargetRadius larger than StreamingMinRadius, or there's no buffer between them.

Model streaming modes
Each Model has a ModelStreamingMode:
- Nonatomic (default): parts stream individually. Set
Workspace.ModelStreamingBehaviorto Improved, which Roblox recommends. - Atomic: the whole model streams in together, so you never see half a model.
- Persistent: sent soon after joining and never streamed out. Use it for things scripts always need.
- PersistentPerPlayer: persistent for chosen players, atomic for others.
The bugs it causes, and fixes
"My script can't find the part." On the client, distant parts may not exist yet. Use WaitForChild on the model, not a part inside it, when the model is Atomic. Wait for Workspace.PersistentLoaded before using Persistent models.
"My changes disappear." Streamed-out instances are parented to nil, not destroyed. Changes a LocalScript made to them can be lost when they stream back in. Keep important state on the server, or reapply it when the part streams in.
"The NPC can't pathfind to it." Client-side pathfinding may fail to reach things that aren't streamed in. Put targets in a Persistent model and wait for PersistentLoaded (pathfinding).
"Distant landmarks vanish." That's streaming working. Use fog to hide the edge, or make landmarks Persistent. Use that sparingly, since persistent content always costs memory.
Is it worth it?
For anything bigger than a small arena, yes. It's one of the most effective ways to make big maps playable on phones (mobile optimisation). For tiny maps where everything is always visible, it adds complexity for little gain.
Lighter assets stream faster
Every mesh and texture streamed to a phone costs time and memory. Vi3W retopologises models from high poly to low poly before export (GLB, FBX or OBJ), so they start lighter. Reuse each mesh and keep textures sensible. Make light props in Vi3W.
Last reviewed

