Add seeking example to explainer / samples

[chcunningham]

In response to user request: "How do we "seek" with the audio decoder?"

The recommendation is

1. call decoder.reset() to drop any pending/queued work
2. call decoder.configure(...) since reset() requires that. if the seek target has the same config as before, simply provide that config again to configure().
3. call decoder.decode(...) with packets starting at the timestamp you're seeking to. Note, you'll need to start with a key frame, but this is mostly not a worry for audio. For video, a typical approach (like used for <video>) would be to start with the key frame at-or-before your seek time, then discard outputs at the output callback until you see the first output with the time you desire.

[tlecoz]

I'm not sure to understand the answer given by chcunningham , but it's not how I proceed

Maybe I didn't understand the correct way to use AudioDecoder but in my case

1. I convert my mp3 into mp4/aac using ffmpeg
2. I use mp4Box to demux the mp4 and I use the data given by mp4Box to produce the encodedAudioChunk I "send" to AudioDecoder
3. AudioDecoder produces AudioFrames (1024 raw data inside for each frame ; as far as I know, you can't change this number)

That's how I work with AudioDecoder right now.

Now, concerning the seek, it's quite simple !
I just use the object "file" produced by mp4Box and call file.seek(time)
Doing that, mp4box with receive the audio (or video) data matching to this time , then - as I said - I use these data to produce some EncodedAudioChunk that I send to AudioDecoder, and AudioDecoder give me the correct raw-data for the time wanted .

(in my case , I ask mp4box to give me a single "sample" each time , it's based on an asynchronous structure but it's so fast it could be considered as synchronous)

[tlecoz]

@chcunningham : As I said to you in an other topic, most of web-developer never heard about muxing/demuxing , it's not a common task for us. I have been "forced" to convert my mp3 into mp4 in order to use mp4box to demux it because I spent so much time to understand how it works, I didn't want to spent weeks of work on the mp3-demuxing part but convert an mp3 into mp4 in order to produce AudioFrame is obviously not the correct way to work.

Can you provide an example that use a mp3-demuxer please

Thanks for reading

[dalecurtis]

You definitely don't need step 1. You can append audio/mpeg frames by configuring the AudioDecoder for mp3. Parsing mp3 is very simple, so I recommend just doing that instead of transmuxing to mp4. https://source.chromium.org/chromium/chromium/src/+/main:media/formats/mpeg/mpeg_audio_stream_parser_base.cc;l=98

[tlecoz]

Hello DaleCurtis
Thank you for your message ! 😀

"Parsing mp3 is very simple"
Ahah, it's very simple for you, it's definitly not a common task for me 😅

Thank you for the code, I will try to translate it into typescript.
If it works I'll post the code here

Thank you again, for everything (I'm fanthomas.lecoz ;) ), and have a nice day !

[hughfenghen]

@tlecoz Hi, :)
Did you succeed in decoding mp3 with AudioDecoder?

[tlecoz]

nope, I used that instead

https://github.com/soundcut/decode-audio-data-fast