diff --git a/README.md b/README.md
index b43adc4b33348c8037330c501a9685b9a17b4d1c..cab2ce74a4a7fde614114c1d203555840c3f44cf 100644
--- a/README.md
+++ b/README.md
@@ -28,9 +28,8 @@ For business inquiries, please visit our website and submit the form: [NVIDIA Re
   - On some machines, `pyexr` refuses to install via `pip`. This can be resolved by installing a pre-built OpenEXR from [here](https://www.lfd.uci.edu/~gohlke/pythonlibs/#openexr).
 - __(optional)__ OptiX __7.3 or higher__ for faster mesh SDF training. Set the environment variable `OptiX_INSTALL_DIR` to the installation directory if it is not discovered automatically.
 
-## Linux
 
-First, install the following packages
+If you are using Linux, we recommend the following packages
 ```sh
 sudo apt-get install build-essential git \
              python3-dev python3-pip libopenexr-dev \
@@ -38,15 +37,15 @@ sudo apt-get install build-essential git \
              libxinerama-dev libxcursor-dev libxi-dev
 ```
 
-Next, we recommend installing CUDA and OptiX in `/usr/local/`.
-Make sure to add your CUDA installation to your path, for example, if you have CUDA 11.4, add the following to your `~/.bashrc`
+We also recommend installing CUDA and OptiX in `/usr/local/` and adding the CUDA installation to your path.
+For example, if you have CUDA 11.4, add the following to your `~/.bashrc`
 ```sh
 export PATH="/usr/local/cuda-11.4/bin:$PATH"
 export LD_LIBRARY_PATH="/usr/local/cuda-11.4/lib64:$LD_LIBRARY_PATH"
 ```
 
 
-# Compilation
+# Compilation (Windows & Linux)
 
 Begin by cloning this repository and all its submodules using the following command:
 ```sh
@@ -72,13 +71,12 @@ If automatic GPU architecture detection fails, (as can happen if you have multip
 
 <img src="docs/assets_readme/testbed.png" width="100%"/>
 
-This codebase comes with an interactive testbed that includes many features beyond our academic publication:
-- Additional training features, such as real-time camera ex- and intrinsics optimization
-- Marching cubes for NeRF->Mesh and SDF->Mesh conversion
-- Various visualization options (e.g. neuron activations)
-- A spline-based camera path editor to create videos
-- Debug visualizations of the activations of every neuron input and output
-- And many more task-specific settings
+This codebase comes with an interactive testbed that includes many features beyond our academic publication, such as:
+- Additional training features, such as real-time camera ex- and intrinsics optimization.
+- Marching cubes for NeRF->Mesh and SDF->Mesh conversion.
+- A spline-based camera path editor to create videos.
+- Debug visualizations of the activations of every neuron input and output.
+- And many more task-specific settings.
 
 
 ## NeRF fox
@@ -89,7 +87,8 @@ One test scene is provided in this repository, using a small number of frames fr
 instant-ngp$ ./build/testbed --scene data/nerf/fox
 ```
 
-Alternatively, download any NeRF-compatible scene (e.g. [from the NeRF authors' drive](https://drive.google.com/drive/folders/1JDdLGDruGNXWnM1eqY1FNL9PlStjaKWi)) into the data subfolder. now you can run:
+Alternatively, download any NeRF-compatible scene (e.g. [from the NeRF authors' drive](https://drive.google.com/drive/folders/1JDdLGDruGNXWnM1eqY1FNL9PlStjaKWi)) into the data subfolder.
+Now you can run:
 
 ```sh
 instant-ngp$ ./build/testbed --scene data/nerf_synthetic/lego
@@ -114,7 +113,7 @@ instant-ngp$ ./build/testbed --scene data/image/albert.exr
 ## Volume Renderer
 
 Download the nanovdb volume file for the Disney Cloud dataset from <a href="https://drive.google.com/drive/folders/1SuycSAOSG64k2KLV7oWgyNWyCvZAkafK?usp=sharing"> this google drive link</a>.
-The dataset is derived from <a href="https://disneyanimation.com/data-sets/?drawer=/resources/clouds/">this</a> dataset which is licensed under a <a href="https://media.disneyanimation.com/uploads/production/data_set_asset/6/asset/License_Cloud.pdf">CC BY-SA 3.0 License</a>.
+The dataset is derived from <a href="https://disneyanimation.com/data-sets/?drawer=/resources/clouds/">this</a> dataset (<a href="https://media.disneyanimation.com/uploads/production/data_set_asset/6/asset/License_Cloud.pdf">CC BY-SA 3.0</a>).
 
 ```sh
 instant-ngp$ ./build/testbed --mode volume --scene data/volume/wdas_cloud_quarter.nvdb
diff --git a/docs/assets/mueller2022instant.pdf b/docs/assets/mueller2022instant.pdf
index 21bfe7c9212da969c61ba51f1a89b52c47061230..3238f3d6083324f113a279578638784fb522e9a4 100644
Binary files a/docs/assets/mueller2022instant.pdf and b/docs/assets/mueller2022instant.pdf differ
diff --git a/docs/index.html b/docs/index.html
index 6fd73b2972c400ea2b0a5807e6b3385b6af53238..acbce2ac7fe7be6dba2adc3f3fc5026ac03e6668 100644
--- a/docs/index.html
+++ b/docs/index.html
@@ -225,7 +225,7 @@ figure {
  <body>
 <div class="container">
 	<div class="paper-title">
-	  <h1>Instant Neural Graphics Primitives with a Multiresolution Hash Encoding</h1>
+		<h1>Instant Neural Graphics Primitives with a Multiresolution Hash Encoding</h1>
 	</div>
 
 	<div id="authors">
@@ -275,7 +275,7 @@ figure {
 
 		<figure style="width: 25%; float: left">
 			<p class="caption_bold">
-				Neural radiance field
+				NeRF
 			</p>
 		</figure>
 
@@ -294,11 +294,11 @@ figure {
 
 		<figure style="width: 100%; float: left">
 			<p class="caption_justify">
-				We demonstrate near-instant training of neural graphics primitives on a single GPU for multiple tasks. In <b>Gigapixel image</b> we represent an image by a neural network. <b>SDF</b> learns a signed distance function in 3D space whose zero level-set represents a 2D surface.
+				We demonstrate near-instant training of neural graphics primitives on a single GPU for multiple tasks. In <b>gigapixel image</b> we represent an image by a neural network. <b>SDF</b> learns a signed distance function in 3D space whose zero level-set represents a 2D surface.
 				<!--<b>Neural radiance caching</b> (NRC) <a href="https://research.nvidia.com/publication/2021-06_Real-time-Neural-Radiance">[Müller et al. 2021]</a> employs a neural network that is trained in real-time to cache costly lighting calculations-->
 				<b>NeRF</b> <a href="https://research.nvidia.com/publication/2021-06_Real-time-Neural-Radiance">[Mildenhall et al. 2020]</a> uses 2D images and their camera poses to reconstruct a volumetric radiance-and-density field that is visualized using ray marching.
-				Lastly, <b>Neural volume</b> learns a denoised radiance and density field directly from a volumetric path tracer.
-				In all tasks, our encoding and its efficient implementation provide clear benefits: near-instant training, high quality, and simplicity. Our encoding is task-agnostic: we use the same implementation and hyperparameters across all tasks and only vary the hash table size which trades off quality and performance.
+				Lastly, <b>neural volume</b> learns a denoised radiance and density field directly from a volumetric path tracer.
+				In all tasks, our encoding and its efficient implementation provide clear benefits: instant training, high quality, and simplicity. Our encoding is task-agnostic: we use the same implementation and hyperparameters across all tasks and only vary the hash table size which trades off quality and performance.
 			</p>
 		</figure>
 	</section>
@@ -334,7 +334,7 @@ figure {
 				Your browser does not support the video tag.
 			</video>
 			<p class="caption">
-				Real-time training progress on the image task where the MLP learns the mapping from 2D coordinates to RGB colors of a high-resolution image. Note that in this video, the network is trained from scratch - but converges so quickly you may miss it if you blink! <br/>
+				Real-time training progress on the image task where the neural network learns the mapping from 2D coordinates to RGB colors of a high-resolution image. Note that in this video, the network is trained from scratch - but converges so quickly you may miss it if you blink! <br/>
 			</p>
 		</figure>
 
@@ -390,7 +390,7 @@ figure {
 			<p class="caption_inline">10k + 12.6M parameters</br>1:45 (mm:ss)</p>
 		</figure>
 		<p class="caption_justify">
-			A demonstration of the reconstruction quality of different encodings and parametric data structures for storing trainable feature embeddings. Each configuration was trained for 11000 steps using our fast NeRF implementation, varying only the input encoding. The number of trainable parameters (MLP weights + encoding parameters) and training time are shown below each image. Our encoding (d) with a similar total number of trainable parameters as the frequency encoding (c) trains over 8 times faster, due to the sparsity of updates to the parameters and smaller MLP. Increasing the number of parameters (e) further improves approximation quality without significantly increasing training time.
+			A demonstration of the reconstruction quality of different encodings. Each configuration was trained for 11000 steps using our fast NeRF implementation, varying only the input encoding and the neural network size. The number of trainable parameters (neural network weights + encoding parameters) and training time are shown below each image. Our encoding <span class="caption_bold">(d)</span> with a similar total number of trainable parameters as the frequency encoding <span class="caption_bold">(c)</span> trains over 8 times faster, due to the sparsity of updates to the parameters and smaller neural network. Increasing the number of parameters <span class="caption_bold">(e)</span> further improves approximation quality without significantly increasing training time.
 		</p>
 		<figure>
 			<video class="centered" width="100%" autoplay muted loop playsinline>
@@ -402,46 +402,43 @@ figure {
 			</p>
 		</figure>
 
-		<figure style="width: 50.0%; float: left">
+		<figure style="width: 49.5%; float: left">
 			<video class="centered" width="100.0%" autoplay muted loop playsinline>
 				<source src="assets/robot.mp4" type="video/mp4">
 				Your browser does not support the video tag.
 			</video>
-			<p class="caption">
-				Fly-through in a trained Neural Radiance Field. Large, natural 360 scenes are well supported.
-			</p>
 		</figure>
 
-		<figure style="width: 50.0%; float: left">
+		<figure style="width: 49.5%; float: right">
 			<video class="centered" width="100.0%" autoplay muted loop playsinline>
 				<source src="assets/modsynth.mp4" type="video/mp4">
 				Your browser does not support the video tag.
 			</video>
-			<p class="caption">
-				Fly-through in a trained Neural Radiance Field.
-				Despite being trained from just 34 photos, this complex scene with many disocclusions and specular surfaces is well reconstructed.
-			</p>
 		</figure>
+		<p class="caption_justify">
+			Fly-throughs of trained real-world NeRFs. Large, natural 360 scenes (left) as well as complex scenes with many disocclusions and specular surfaces (right) are well supported.
+			Both models can be rendered in real time and were trained in under 5 minutes from casually captured data: the left one from an iPhone video and the right one from 34 photographs.
+		</p>
 
-		<figure style="width: 50.0%;">
+		<figure style="width: 75.0%;">
 			<video class="centered" width="100.0%" autoplay muted loop playsinline>
 				<source src="assets/cloud_training.mp4" type="video/mp4">
 				Your browser does not support the video tag.
 			</video>
 			<p class="caption">
-				We train NeRF-like radiance fields from the noisy output of a volumetric path tracer. Rays are fed in real-time to the network during training, which learns a denoised radiance field.
+				We also support training NeRF-like radiance fields from the noisy output of a volumetric path tracer. Rays are fed in real-time to the network during training, which learns a denoised radiance field.
 			</p>
 		</figure>
 
 		<h3>Signed Distance Function</h3>
 		<hr>
-		<figure style="width: 50.0%;">
+		<figure style="width: 100.0%;">
 			<video class="centered" width="100.0%" autoplay muted loop playsinline>
 				<source src="assets/sdf_grid_lq.mp4" type="video/mp4">
 				Your browser does not support the video tag.
 			</video>
 			<p class="caption">
-				Real-time training progress on various SDF datsets. Training data is generated on the fly from the ground-truth mesh using the NVIDIA OptiX raytracing framework.
+				Real-time training progress on various SDF datsets. Training data is generated on the fly from the ground-truth mesh using the <a href="https://developer.nvidia.com/optix">NVIDIA OptiX raytracing framework</a>.
 			</p>
 		</figure>
 
@@ -453,7 +450,7 @@ figure {
 				Your browser does not support the video tag.
 			</video>
 			<p class="caption">
-				Direct visualization of a Neural Radiance Cache, in which the network preducts outgoing radiance at the first non-specular vertex of each pixel's path, and is trained on-line from rays generated by a realtime path-tracer. On the left, we show results using the triangular encoding of <a href="https://research.nvidia.com/publication/2021-06_Real-time-Neural-Radiance">[Müller et al. 2021]</a>; on the right, the new Multiresolution Hash Encoding allows the network to learn much sharper details, for example in the shadow regions.
+				Direct visualization of a <em>neural radiance cache</em>, in which the network predicts outgoing radiance at the first non-specular vertex of each pixel's path, and is trained on-line from rays generated by a real-time pathtracer. On the left, we show results using the triangle wave encoding of <a href="https://research.nvidia.com/publication/2021-06_Real-time-Neural-Radiance">[Müller et al. 2021]</a>; on the right, the new multiresolution hash encoding allows the network to learn much sharper details, for example in the shadow regions.
 			</p>
 		</figure>
 
@@ -510,11 +507,13 @@ figure {
 			<a href="https://www.cs.toronto.edu/~jlucas/">James Lucas</a> and
 			<a href="https://tovacinni.github.io">Towaki Takikawa</a>
 			for proof-reading and feedback.
-			We also thank <a href="https://tovacinni.github.io">Towaki Takikawa</a> for providing us with the framework for this website.
+			We also thank <a href="https://joeylitalien.github.io/">Joey Litalien</a> for providing us with the framework for this website.
 			<br/>
 			<em>Girl With a Pearl Earing</em> renovation by Koorosh Orooj <a href="http://profoundism.com/free_licenses.html">(CC BY-SA 4.0 License)</a>
 			<br/>
 			<em>Lucy</em> model from the <a href="http://graphics.stanford.edu/data/3Dscanrep/">Stanford 3D scan repository</a>
+			<br/>
+			<em>Disney Cloud</em> model by Walt Disney Animation Studios. (<a href="https://media.disneyanimation.com/uploads/production/data_set_asset/6/asset/License_Cloud.pdf">CC BY-SA 3.0</a>)
 			</p>
 		</div>
 	</section>
diff --git a/scripts/common.py b/scripts/common.py
index 4595034a71c6fe886b6c3eb4b8d0d2be3cc9471f..85b406bedbd7043b157b8905be67360bf81afa14 100644
--- a/scripts/common.py
+++ b/scripts/common.py
@@ -38,7 +38,8 @@ DATA_FOLDER = SCRIPTS_FOLDER/"data"
 ROOT_DIR = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
 RESULTS_DIR = os.path.join(ROOT_DIR, "results")
 NGP_DATA_FOLDER = os.environ.get("NGP_DATA_FOLDER") or os.path.join(ROOT_DIR, "data")
-#print(f"NGP_DATA_FOLDER is {NGP_DATA_FOLDER}")
+
+
 NERF_DATA_FOLDER = os.path.join(NGP_DATA_FOLDER, "nerf")
 SDF_DATA_FOLDER = os.path.join(NGP_DATA_FOLDER, "sdf")
 IMAGE_DATA_FOLDER = os.path.join(NGP_DATA_FOLDER, "image")