Local AI Runtime
Learn how to tune RenamerX local AI runtime settings for memory usage, GPU policy, parallel slots, context size, and image detail.
Local AI Runtime
RenamerX uses a local AI runtime, so runtime settings influence quality, speed, and resource efficiency.
Most users should leave the defaults alone. This page is for the cases where you need to tune performance or stability.
What RenamerX installs locally
After first launch, RenamerX prepares the local resources needed for on-device AI processing and media analysis, including:
- A local AI processing engine
- Model resources for document and image understanding
- Media-analysis components used by video workflows
Runtime settings in Preferences
Available runtime controls include:
- Performance mode
- Context size
- GPU policy
- Parallel slots
- Image detail
Most of these changes are intended to take effect on the next app start, so avoid expecting every runtime adjustment to reconfigure the active session immediately.
Performance mode
RenamerX includes three performance profiles:
- lowMemory
- balanced
- highPerformance
These profiles change multiple settings together.
Low Memory
Works well on smaller machines or when you want a lower resource footprint.
Balanced
The default profile. This is the right starting point for most users.
High Performance
Best when you have stronger hardware and want more throughput. It uses more aggressive runtime settings.
Context size
Available context sizes are:
- 4096
- 8192
- 12288
Higher context can help with more complex extraction, but it also increases memory pressure.
GPU policy
Available GPU policies are:
- auto
- aggressive
- cpuOnly
Use:
- auto for most systems
- aggressive when you know your machine handles GPU inference well
- cpuOnly when GPU inference causes instability or you want predictable CPU-only behavior
Parallel slots
Available parallel settings are:
- 1 slot
- 2 slots
Higher parallelism can improve throughput, but it can also increase RAM and GPU pressure.
Image detail
Available image-detail levels are:
- fast
- balanced
- detailed
Use detailed when visual nuance matters more than throughput. Use fast when you are optimizing for speed.
Recommended tuning patterns
If the app feels heavy
- Switch to Low Memory
- Reduce context size
- Use CPU Only
- Keep parallel slots at 1
If image and video naming feels weak
- Keep Balanced or High Performance
- Increase image detail
- Make sure local resources finished downloading completely
If you are processing mixed media all day
- Stay on Balanced first
- Increase only one setting at a time
- Measure stability before enabling more parallel work
A practical rule
Do not tune everything at once. Change one runtime setting, run a real batch, and judge the result with your own files.