diff --git a/README.md b/README.md
index 65bcdc4a..876a7d07 100644
--- a/README.md
+++ b/README.md
@@ -72,21 +72,7 @@ uvx -p 3.13 "git+https://github.com/fdrgsp/cali[cp4]"
 
 ***NOTE**: use python 3.11 or greater. Swap `cp4` with `cp3` to use Cellpose 3.x.*
 
-<br>
-
-To use Cellpose with GPU (e.g. on Windows), the correct version of CUDA has to be installed:
-
-```bash
-uvx -p 3.13 --index pytorch-cu130=https://download.pytorch.org/whl/cu130 "git+https://github.com/fdrgsp/cali[cp4,cu130]"
-```
-
-***NOTE**: swap `cu130` with `cu126` or `cu128` for different CUDA versions and `cp4` with `cp3` to use Cellpose 3.x.*
-
-***NOTE**: [NVIDIA Drivers](https://www.nvidia.com/en-us/drivers/) should be already installed. You can check what version of cuda you need by running:*
-
-```bash
-nvidia-smi
-```
+***NOTE**: `uvx` supports GPU acceleration on **macOS only** (via Apple MPS — no CUDA required). On Windows/Linux, use `(uv) pip install` instead (see [Installation](#installation)).*
 
 ### Installation
 
@@ -102,19 +88,15 @@ Install `cali` directly without cloning (use python 3.11 or greater):
 
 <br>
 
-To use Cellpose with GPU (e.g. on Windows), the correct version of CUDA has to be installed:
+To use Cellpose with GPU on Windows/Linux, install PyTorch with the CUDA version matching your driver as a second step:
 
 ```bash
-(uv) pip install --index pytorch-cu130=https://download.pytorch.org/whl/cu130 "git+https://github.com/fdrgsp/cali[cp4,cu130]"
+(uv) pip install torch torchvision --index-url https://download.pytorch.org/whl/cu{VERSION}
 ```
 
-***NOTE**: swap `cu130` with `cu126` or `cu128` for different CUDA versions and `cp4` with `cp3` to use Cellpose 3.x.*
+Replace `{VERSION}` with your CUDA version (e.g. `cu126`, `cu128`, `cu130`). Run `nvidia-smi` to check which version your driver supports.
 
-***NOTE**: [NVIDIA Drivers](https://www.nvidia.com/en-us/drivers/) should be already installed. You can check what version of cuda you need by running:*
-
-```bash
-nvidia-smi
-```
+***NOTE**: [NVIDIA Drivers](https://www.nvidia.com/en-us/drivers/) should be already installed.*
 
 #### ▶ Using uv
 
@@ -135,19 +117,15 @@ uv sync --extra cp4
 
 <br>
 
-To use Cellpose with GPU (e.g. on Windows), the correct version of CUDA has to be installed. Add a CUDA extra alongside your Cellpose extra:
+To use Cellpose with GPU on Windows/Linux, install PyTorch with the CUDA version matching your driver as a second step:
 
 ```bash
-uv sync --extra cp4 --extra cu130   # Cellpose 4.x + CUDA 13.0
+uv pip install torch torchvision --index-url https://download.pytorch.org/whl/cu{VERSION}
 ```
 
-***NOTE**: swap `cu130` with `cu126` or `cu128` for different CUDA versions.*
-
-***NOTE**: [NVIDIA Drivers](https://www.nvidia.com/en-us/drivers/) should be already installed. You can check what version of cuda you need by running:*
+Replace `{VERSION}` with your CUDA version (e.g. `cu126`, `cu128`, `cu130`). Run `nvidia-smi` to check which version your driver supports.
 
-```bash
-nvidia-smi
-```
+***NOTE**: [NVIDIA Drivers](https://www.nvidia.com/en-us/drivers/) should be already installed.*
 
 #### ▶ NOTES
 
diff --git a/pyproject.toml b/pyproject.toml
index 56d25163..1ae0ef4c 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -61,10 +61,6 @@ dependencies = [
 # Cellpose is optional - choose the version that fits your Python version
 cp3 = ["cellpose>=3.0.0,<4.0.0"]
 cp4 = ["cellpose>=4.0.0"]
-# CUDA variants for GPU-accelerated PyTorch (Windows/Linux only)
-cu126 = ["torch", "torchvision"]
-cu128 = ["torch", "torchvision"]
-cu130 = ["torch", "torchvision"]
 
 [tool.uv]
 conflicts = [
@@ -72,11 +68,6 @@ conflicts = [
       { extra = "cp3" },
       { extra = "cp4" },
     ],
-    [
-      { extra = "cu126" },
-      { extra = "cu128" },
-      { extra = "cu130" },
-    ],
 ]
 
 [dependency-groups]
@@ -102,31 +93,6 @@ cali = "cali.__main__:main"
 
 [tool.uv.sources]
 cali = { workspace = true }
-torch = [
-    { index = "pytorch-cu126", extra = "cu126" },
-    { index = "pytorch-cu128", extra = "cu128" },
-    { index = "pytorch-cu130", extra = "cu130" },
-]
-torchvision = [
-    { index = "pytorch-cu126", extra = "cu126" },
-    { index = "pytorch-cu128", extra = "cu128" },
-    { index = "pytorch-cu130", extra = "cu130" },
-]
-
-[[tool.uv.index]]
-name = "pytorch-cu126"
-url = "https://download.pytorch.org/whl/cu126"
-explicit = true
-
-[[tool.uv.index]]
-name = "pytorch-cu128"
-url = "https://download.pytorch.org/whl/cu128"
-explicit = true
-
-[[tool.uv.index]]
-name = "pytorch-cu130"
-url = "https://download.pytorch.org/whl/cu130"
-explicit = true
 
 [tool.hatch.build.targets.wheel]
 only-include = ["src"]