MediaRecorder API example

MediaRecorder API example. Record a MediaStream with a chosen codec, then download the Blob. A stream is not a file.

The problem

A live MediaStream is pixels in motion. Saving a video means encoding those frames into a container the browser can play later.

The method

MediaRecorder is the high-level encoder. Hand it a stream, get a file.

VideoEncoder is the other path. You pull frames, you compress them, you get packets. There is no file until something muxes those packets. This lab skips the muxer and unpacks the packets with VideoDecoder onto a canvas.

What to do

  1. Run isTypeSupported(). Pick a mime type.
  2. Start the camera. Record with MediaRecorder.
  3. Capture a batch and encode it with each mime. Same frames, different codecs.
  4. Record with VideoEncoder and watch the frame count.

MediaRecorder.isTypeSupported()

What this browser will encode. The selected type is what MediaRecorder gets.

MediaRecorder.isTypeSupported("video/webm;codecs=vp8")
MediaRecorder.isTypeSupported("video/mp4;codecs=avc1")

Pick a mime type.

MediaRecorder

Takes a MediaStream and writes a Blob. start(1000) emits a chunk every second. Pause, resume, and requestData() work while it runs. Video only in this lab. Add an audio track and it will mux that too.

const rec = new MediaRecorder(stream, { mimeType })
rec.ondataavailable = (e) => chunks.push(e.data)
rec.start(1000)
rec.stop()
const blob = new Blob(chunks, { type: rec.mimeType })

Same frames, every codec

Captures frames from the preview, then MediaRecorder encodes that same batch with any mime you pick.

frames.push(await createImageBitmap(video))

const stream = canvas.captureStream(0)
const rec = new MediaRecorder(stream, { mimeType })
rec.start()
const track = stream.getVideoTracks()[0]
for (const frame of frames) {
  ctx.drawImage(frame, 0, 0)
  track.requestFrame()
  await new Promise((resolve) => setTimeout(resolve, 1000 / 30))
}
rec.stop()

VideoEncoder

Yes

MediaRecorder hides this loop. Here you run it yourself. A cloned track keeps the preview alive. Each VideoFrame is compressed into an EncodedVideoChunk. That is a packet, not a movie. Stop to run the packets backwards through VideoDecoder.

  1. Clone the camera track.
  2. MediaStreamTrackProcessor pulls one VideoFrame at a time.
  3. encoder.encode(frame) compresses it. output receives a chunk.
  4. Stop. VideoDecoder draws those chunks on a canvas. Still not a file.
const processor = new MediaStreamTrackProcessor({ track: track.clone() })
const encoder = new VideoEncoder({
  output: (chunk) => chunks.push(chunk),
  error,
})
encoder.configure({ codec: "vp8", width, height, bitrate: 2e6 })

const { value: frame } = await processor.readable.getReader().read()
encoder.encode(frame)
frame.close()

All Web API labs

  • HTML usermedia element example
  • Web Audio API example
  • Web Speech API example