System requirements
Short answer: Build Perch runs on 64-bit Windows 11 with a 4-core x64 CPU, 8 GB of RAM, and about 1 GB of free disk space; no graphics card needed. A 6-core Intel Core i5 (10th-gen or newer) or comparable AMD Ryzen keeps local transcription, workspace indexing, and video editing smooth.
Full spec sheet
| Requirement | Details |
|---|---|
| Operating system | Windows 11, 64-bit (x64). Windows on Arm is not supported. |
| Processor | Minimum: 4-core x64 CPU with AVX2: any mainstream Intel Core i3/i5 or AMD Ryzen 3/5. Recommended: a 6-core-or-more modern CPU (Intel Core i5 10th-gen or newer, or a comparable AMD Ryzen) for smooth local transcription, indexing large workspaces, and local video editing. |
| Memory | 8 GB RAM; 16 GB recommended if running local models. |
| Graphics | None required by Build Perch. Models run through local Ollama have their own GPU requirements. |
| Storage | About 1 GB free: the app installs at ~420 MB with speech, embedding, and media models included, and the local search index grows with your workspace. An SSD is recommended. |
| Internet | Required for trial start, license activation, and cloud models. Runs up to 30 days offline after activation; local models work fully offline. |
| Model access | One of: an OpenRouter API key (cloud models), Ollama (free local models), or an Ollama Cloud subscription. |
| Local models (optional) | Separate Ollama install, auto-detected. About 8 GB RAM for 7–8B models; a dedicated GPU with matching VRAM for larger ones. |
| Dependencies | WebView2 Runtime (built into Windows 11; the installer runs Microsoft’s bootstrapper if it’s somehow missing) and the Microsoft Visual C++ Redistributable. |
Do I need a graphics card?
No. Every built-in AI feature (speech-to-text and local search embeddings) runs on the CPU only; the app never asks for a GPU. A GPU only matters if you additionally run large models through a local Ollama install, and that’s an optional, separate setup.
What if my CPU doesn’t have AVX2?
Voice dictation and local transcription turn off with a clear in-app message instead of crashing; everything else keeps working. Those two features need a CPU with AVX2, which has been standard on Intel Core chips since 2013 and on every AMD Ryzen; the only exceptions are a handful of low-power Celeron, Pentium Silver, and Atom-class chips. Local search, local media editing, and the rest of the app run on any x64 processor, though those same budget chips are slow at on-device work like workspace indexing and video editing, so they aren’t the best fit if you lean on the local AI features.
Does the app download anything on first run?
No. The local speech-to-text and search models ship inside the installer, along with everything needed for local media editing. Nothing downloads the first time you launch the app.
Do I need an internet connection?
Only for a few specific things: starting the trial, activating a license, and talking to cloud models (OpenRouter or Ollama Cloud). After activation, the app keeps working for up to 30 days without a connection. Local Ollama models, local search, voice dictation, and local media editing all work fully offline regardless.
What about running local models through Ollama?
Ollama is a separate, free install the app detects automatically; Build Perch doesn’t bundle it. Sizing for local models is Ollama’s, not Build Perch’s: as a rough guide, plan on about 8 GB of RAM for 7–8B parameter models, and a GPU with enough VRAM for anything larger.
Good to know
- The installer is a standard
.exe(BuildPerch_<version>_x64-setup.exe); there’s no MSI or MSIX package. - Build Perch supports Windows 11 only; Windows 10 isn’t supported.