Post by Usagi Lab in AI Face VTuber – turn one photo into a real-time avatar (free trial)

Viewing post in AI Face VTuber – turn one photo into a real-time avatar (free trial)

Thank you so much for the detailed report - the log snippet you included is

genuinely helpful.

Based on what you shared, this looks like a compatibility issue between the

app's TensorRT/CUDA build step and your GPU. The RTX 5060 Ti is a very new

(Blackwell) architecture, and the "AttributeError: context" you're seeing is

actually a side effect, not the real cause - it happens because engine

creation for the WarpingSpade model fails earlier, before that object is

fully initialized.

To confirm and fix this, could you share:

1. Your full app_log.txt (not just the tail - the earlier lines around the

WarpingSpade build will have the actual error)

2. Your NVIDIA driver version (nvidia-smi or GeForce Experience)

I'm looking into adding proper support for RTX 50-series GPUs and will

follow up here once I have an update. Really sorry for the trouble, and

thanks for your patience!

Thanks for looking into this! I deleted everything but my GPU/driver information is:

GPU: NVIDIA GeForce RTX 5060 Ti 16GB NVIDIA Driver: 591.86 CUDA Version reported by nvidia-smi: 13.1

Thanks, that's exactly what I needed - your driver (591.86) is new enough,

so the problem is on the app's side.

I confirmed the cause: this build bundles TensorRT 9.0 and PyTorch with

CUDA 12.1, which predate the RTX 50 series (Blackwell). Those need CUDA

12.8+ and newer TensorRT, so the app currently can't run on your RTX 5060 Ti.

No need to send the log anymore.

I'm evaluating a separate build for RTX 50-series GPUs, but it needs a

full upgrade of the bundled libraries, so I can't promise a date yet. I'm

also on an older GPU here, so I can't test Blackwell myself. If I get a

build working, would you be willing to try it as a beta tester? I'll

post here either way.

I'll also add a note about supported GPUs (RTX 20/30/40) to the store

page. Sorry for the trouble, and thank you for reporting it so clearly!