MinimumRuns
- System
- macOS 26 or later
- Processor
- Apple M1
- Memory
- 8 GB unified memory
- Graphics
- Built-in GPU
- Free space
- 20 GB, SSD
- Runs well
- Templates and lessons with the small writing model; slower renders
System requirements
What your computer needs.
Local AI uses your computer’s memory and graphics hardware. Use these configurations as guidance; speed and model support depend on your setup.
Mac
Apple silicon shares one pool of memory between the processor and the graphics, so memory is the number to look at.
RecommendedBest value
- System
- macOS 26 or later
- Processor
- Apple M2 Pro, M3 or M4
- Memory
- 16 GB unified memory
- Graphics
- Built-in GPU, 10 cores or more
- Free space
- 50 GB, SSD
- Runs well
- Every mode, the most natural Arabic and French voice, and the mid-size AI Generate model
BestMore demanding models
- System
- macOS 26 or later
- Processor
- Apple M3 Max, M4 Max or Ultra
- Memory
- 36 GB unified memory or more
- Graphics
- Built-in GPU, 30 cores or more
- Free space
- 100 GB, SSD
- Runs well
- Every AI model, including the largest AI Generate model, with fast 4K renders
Windows
On a PC the graphics card holds the AI model, so its video memory (VRAM) matters more than its speed. NVIDIA cards work best.
MinimumRuns
- System
- Windows 10 or 11, 64-bit
- Processor
- 6 cores (Ryzen 5 5600 or Core i5-12400)
- Memory
- 16 GB
- Graphics
- NVIDIA with 6 GB VRAM (RTX 2060 or 3050); runs without one, slowly
- Free space
- 30 GB, SSD
- Runs well
- Templates and lessons with the standard writing model
RecommendedBest value
- System
- Windows 11, 64-bit
- Processor
- 8 cores (Ryzen 7 7700 or Core i7-13700)
- Memory
- 32 GB
- Graphics
- NVIDIA RTX 3060 12 GB or RTX 4060 Ti 16 GB
- Free space
- 100 GB, NVMe SSD
- Runs well
- Every mode with the larger writing model and the mid-size AI Generate model
BestMore demanding models
- System
- Windows 11, 64-bit
- Processor
- Ryzen 9 or Core i9
- Memory
- 64 GB
- Graphics
- NVIDIA RTX 4090 or 5090, 24 GB VRAM or more
- Free space
- 200 GB, NVMe SSD
- Runs well
- Every AI model, including the largest AI Generate model, with fast 4K renders
LinuxComing soon
The same hardware as Windows. Use the NVIDIA proprietary driver so the AI models run on the graphics card.
MinimumRuns
- System
- Ubuntu 22.04, Debian 12 or later, 64-bit
- Processor
- 6 cores (Ryzen 5 5600 or Core i5-12400)
- Memory
- 16 GB
- Graphics
- NVIDIA with 6 GB VRAM; runs without one, slowly
- Free space
- 30 GB, SSD
- Runs well
- Templates and lessons with the standard writing model
RecommendedBest value
- System
- Ubuntu 24.04, 64-bit
- Processor
- 8 cores (Ryzen 7 7700 or Core i7-13700)
- Memory
- 32 GB
- Graphics
- NVIDIA RTX 3060 12 GB or RTX 4060 Ti 16 GB
- Free space
- 100 GB, NVMe SSD
- Runs well
- Every mode with the larger writing model and the mid-size AI Generate model
BestMore demanding models
- System
- Ubuntu 24.04, 64-bit
- Processor
- Ryzen 9 or Core i9
- Memory
- 64 GB
- Graphics
- NVIDIA RTX 4090 or 5090, 24 GB VRAM or more
- Free space
- 200 GB, NVMe SSD
- Runs well
- Every AI model, including the largest AI Generate model, with fast 4K renders
Choosing your setup
Memory first
The AI models have to fit in memory. A 12 GB graphics card is a better choice for Movyra than a faster card with 8 GB.
Cores for rendering
Rendering draws several frames at once on the processor, so more cores means shorter exports.
Keep space free
Models are downloaded the first time you use them, from under 1 GB up to about 19 GB for the largest.