Add missing stuff in docs

docs/guide/packets-and-samples.md

@@ -274,6 +274,26 @@ const sample = new VideoSample(buffer, {
 
 See [`VideoPixelFormat`](https://w3c.github.io/webcodecs/#enumdef-videopixelformat) for a list of pixel formats supported by WebCodecs.
 
+#### Custom resource constructor
+
+For advanced use cases (custom decoders, GPU-backed frames, etc.), you can back a `VideoSample` with your own implementation of [`VideoSampleResource`](../api/VideoSampleResource):
+
+```ts
+import { VideoSample, VideoSampleResource } from 'mediabunny';
+
+class MyResource extends VideoSampleResource {
+	// Implement getFormat(), getCodedWidth(), getCodedHeight(),
+	// getSquarePixelWidth(), getSquarePixelHeight(), getColorSpace(),
+	// getDataPlanes(), toRgbSample(), and close().
+}
+
+const sample = new VideoSample(new MyResource(), {
+	timestamp: 0,
+});
+```
+
+This allows you to back a `VideoSample` with your own data without having to copy it.
+
 ### Inspecting video samples
 
 A `VideoSample` has several read-only properties:
@@ -402,6 +422,31 @@ You can pass additional options to `allocationSize` and `copyTo` to extract data
 
 ---
 
+You can transform a `VideoSample` to resize, rotate, and/or crop it, producing a new `VideoSample`:
+```ts
+const transformed = await videoSample.transform({
+	width: 640,
+	height: 360,
+	fit: 'cover',
+	rotate: 90,
+	crop: { left: 0, top: 0, width: 1920, height: 1000 },
+	alpha: 'discard',
+});
+```
+
+In browser environments, the transform is performed using a canvas. In non-browser environments without `OffscreenCanvas` or `HTMLCanvasElement`, this method throws unless you register a custom transformer:
+```ts
+import { registerVideoSampleTransformer } from 'mediabunny';
+
+registerVideoSampleTransformer((sample, description) => {
+	// Return a transformed VideoSample, or null to defer to the next transformer.
+});
+```
+
+The [`@mediabunny/server`](./extensions/server) extension registers such a transformer.
+
+---
+
 You can also clone a `VideoSample`:
 ```ts
 const clonedSample = videoSample.clone(); // => VideoSample
@@ -463,6 +508,23 @@ Planar formats store each channel's data contiguously, while interleaved formats
 
 ![Planar vs. interleaved formats](../assets/planar_interleaved.svg)
 
+#### Custom resource constructor
+
+For advanced use cases (custom decoders, etc.), you can back an `AudioSample` with your own implementation of [`AudioSampleResource`](../api/AudioSampleResource):
+
+```ts
+import { AudioSample, AudioSampleResource } from 'mediabunny';
+
+class MyResource extends AudioSampleResource {
+	// Implement getFormat(), getSampleRate(), getNumberOfFrames(),
+	// getNumberOfChannels(), getTimestamp(), getDataPlane(), and close().
+}
+
+const sample = new AudioSample(new MyResource());
+```
+
+This allows you to back an `AudioSample` with your own data without having to copy it.
+
 ### Inspecting audio samples
 
 An `AudioSample` has several read-only properties: