Turn your own documents and YouTube transcripts into a private, offline chatbot — accelerated locally by your GeForce RTX GPU.
Download from NVIDIAOpens NVIDIA's official ChatRTX page · ~35–40 GB download
Publisher
NVIDIA
Cost
Free
GPU needed
RTX 30/40-series
Download size
~35–40 GB
Point it at a folder of .txt, .pdf, .doc/.docx, or .xml files — it indexes them in seconds.
Drop in a video or playlist URL and ask questions about its content, processed locally.
Uses OpenAI's CLIP model to search your own photos by describing what's in them.
Query by speaking, powered by a local Whisper model — no cloud round-trip.
| GPU | GeForce RTX 30/40-series or RTX Ampere/Ada workstation card, 8 GB+ VRAM |
| OS | Windows 10 or 11 |
| RAM | 16 GB or more |
| Driver | 535.11 or newer |
| Storage | ~100 GB free recommended (download + extracted models) |
Click the download button above
It goes straight to NVIDIA's own ChatRTX page — check the requirements listed there match your GPU before downloading the large file.
Extract the downloaded package
It's a large compressed file (~35 GB) — extracting it needs roughly that much free space again.
Run setup.exe
Keep the default install path unless you have a specific reason to change it — some users report issues after moving it.
Let it finish setting up on first launch
A command prompt window runs in the background configuring things; leave it be until your browser opens the chat interface automatically.