diff --git a/README.md b/README.md
index 457c1b8a..f854cee3 100644
--- a/README.md
+++ b/README.md
@@ -16,8 +16,9 @@ _NOTE: This only works on desktop Chrome 129+ currently (Oct 2024). Firefox and
 ## Features
 
 The demo can load pretrained ply splats, and can load datasets to train on. Currently only two formats are supported. A .zip file containing:
-- A transforms_train.json and images, like the synthetic nerf scene dataset.
 - An `images` & `sparse` folder with [`COLMAP`](https://github.com/colmap/colmap) data
+- A .json and images, like the [nerfstudio format](https://docs.nerf.studio/quickstart/data_conventions.html).
+  - You can specify a custom transforms_train.json and transforms_eval.json split.
 
 While training you can interact with the scene and see the training dynamics live, and compare the current rendering to training / eval views as the training progresses.
 
@@ -72,7 +73,7 @@ Brush is split into various crates. A quick overview of the different responsibi
 - `brush-viewer` handles the UI and integrating the training loop.
 - `brush-android` is the binary target for running on android, while `brush-desktop` is for running both on web, and mac/Windows/Linux.
 - `brush-wgsl` handles some kernel inspection for generating CPU-side structs and interacing with [naga-oil](https://github.com/bevyengine/naga_oil) to handle shader imports.
-- `brush-dataset` handles importing different datasets like COLMAP or synthetic nerf data.
+- `brush-dataset` handles importing different training data formats.
 - `brush-prefix-sum` and `brush-sort` are only compute kernels and should be largely independent of Brush (other than `brush-wgsl`).
 - `rrfd` is a small extension of [`rfd`](https://github.com/PolyMeilex/rfd)
 
diff --git a/crates/brush-dataset/src/formats/mod.rs b/crates/brush-dataset/src/formats/mod.rs
index 701b0fbc..2a953da3 100644
--- a/crates/brush-dataset/src/formats/mod.rs
+++ b/crates/brush-dataset/src/formats/mod.rs
@@ -15,10 +15,8 @@ pub fn load_dataset<B: Backend>(
     load_args: &LoadDatasetArgs,
     device: &B::Device,
 ) -> anyhow::Result<(DataStream<Splats<B>>, DataStream<Dataset>)> {
-    let streams = nerfstudio::read_dataset(archive.clone(), load_args, device).or_else(|e| {
-        log::info!("Not a NeRF synthetic dataset ({e}), trying to load as Colmap.");
-        colmap::load_dataset::<B>(archive.clone(), load_args, device)
-    });
+    let streams = nerfstudio::read_dataset(archive.clone(), load_args, device)
+        .or_else(|_| colmap::load_dataset::<B>(archive.clone(), load_args, device));
 
     let Ok(streams) = streams else {
         anyhow::bail!("Couldn't parse dataset as any format. Only some formats are supported.")
diff --git a/crates/brush-dataset/src/formats/nerfstudio.rs b/crates/brush-dataset/src/formats/nerfstudio.rs
index f18985d7..f2e2d269 100644
--- a/crates/brush-dataset/src/formats/nerfstudio.rs
+++ b/crates/brush-dataset/src/formats/nerfstudio.rs
@@ -17,7 +17,7 @@ use tokio_stream::StreamExt;
 
 #[derive(serde::Deserialize, Clone)]
 #[allow(unused)] // not reading camera distortions yet.
-struct SyntheticScene {
+struct JsonScene {
     // Simple synthetic nerf camera model.
     camera_angle_x: Option<f64>,
     // Not really used atm.
@@ -92,7 +92,7 @@ struct FrameData {
 }
 
 fn read_transforms_file(
-    scene: SyntheticScene,
+    scene: JsonScene,
     transforms_path: PathBuf,
     archive: DatasetZip,
     load_args: &LoadDatasetArgs,
@@ -175,11 +175,10 @@ pub fn read_dataset<B: Backend>(
     load_args: &LoadDatasetArgs,
     device: &B::Device,
 ) -> Result<(DataStream<Splats<B>>, DataStream<Dataset>)> {
-    log::info!("Loading nerf synthetic dataset");
+    log::info!("Loading nerfstudio dataset");
 
     let transforms_path = archive.find_with_extension(".json", "_train")?;
-    let train_scene: SyntheticScene =
-        serde_json::from_reader(archive.file_at_path(&transforms_path)?)?;
+    let train_scene: JsonScene = serde_json::from_reader(archive.file_at_path(&transforms_path)?)?;
     let train_handles = read_transforms_file(
         train_scene.clone(),
         transforms_path.clone(),
diff --git a/crates/brush-viewer/src/panels/scene.rs b/crates/brush-viewer/src/panels/scene.rs
index b29b29ad..2612642c 100644
--- a/crates/brush-viewer/src/panels/scene.rs
+++ b/crates/brush-viewer/src/panels/scene.rs
@@ -198,7 +198,7 @@ impl ViewerPanel for ScenePanel {
 Load a pretrained .ply file to view it
 
 Or load a dataset to train on. These are zip files with:
-    - a transforms_train.json and images, like the synthetic NeRF dataset format.
+    - a transforms.json and images, like the nerfstudio dataset format.
     - COLMAP data, containing the `images` & `sparse` folder."#,
             );