2x and 4x oversampling for WaveShaperNode #332
Conversation
khodzha
force-pushed
the
oversampling_speex
branch
4 times, most recently
from
0b45b68
to
6aac50e
Compare
audio/Cargo.toml
Outdated
| @@ -18,6 +18,10 @@ servo-media-player = { path = "../player" } | |||
| servo-media-traits = { path = "../traits" } | |||
| smallvec = "0.6.1" | |||
|
|
|||
| [dependencies.speexdsp] | |||
| git = "https://github.com/rust-av/speexdsp-rs" | |||
There was a problem hiding this comment.
There was a problem hiding this comment.
Looks like they're willing to make one soon! rust-av/speexdsp-rs#10
audio/wave_shaper_node.rs
Outdated
| fn apply_curve(buf: &mut [f32], curve: &Vec<f32>) { | ||
| buf.iter_mut().for_each(|sample| { | ||
| let len = curve.len(); | ||
| let curve_index: f32 = ((len - 1) as f32) * (*sample + 1.) / 2.; |
There was a problem hiding this comment.
worth moving (len - 1) / 2 and len out of the closure. Probably will be done by the compiler anyway, but doesn't hurt.
There was a problem hiding this comment.
agreed, I'll fix it up tomorrow
khodzha
force-pushed
the
oversampling_speex
branch
2 times, most recently
from
182bfdd
to
7acd6ca
Compare
|
I prepared a crate with the stand alone resampler, if you could please confirm that the API exposed is fine for you I'll publish it. |
|
@khodzha ^^ |
khodzha
force-pushed
the
oversampling_speex
branch
2 times, most recently
from
9ab1116
to
79094ce
Compare
|
Updated PR to use this crate, API seems ok to me 🤷♂️ |
|
https://crates.io/crates/speexdsp-resampler published ! |
audio/wave_shaper_node.rs
Outdated
| @@ -79,33 +93,68 @@ impl AudioNodeEngine for WaveShaperNode { | |||
| AudioNodeType::WaveShaperNode | |||
| } | |||
|
|
|||
| fn process(&mut self, mut inputs: Chunk, _info: &BlockInfo) -> Chunk { | |||
| fn process(&mut self, mut inputs: Chunk, info: &BlockInfo) -> Chunk { | |||
| debug_assert!(inputs.len() == 1); | |||
|
|
|||
| if inputs.blocks[0].is_silence() { | |||
There was a problem hiding this comment.
@Manishearth could you confirm whether tail-time here could be implemented like this:
if we hit two silent blocks in row after non-silent block, we explicit_silence() them and run them through the waveshaping so all the tail-time from resampler would be spitted out
khodzha
force-pushed
the
oversampling_speex
branch
from
79094ce
to
984bc8f
Compare
|
Added tail-time handling and properly handled the case when waveshaped silence is nonsilent |
khodzha
force-pushed
the
oversampling_speex
branch
from
984bc8f
to
1ab5c50
Compare
khodzha
force-pushed
the
oversampling_speex
branch
from
1ab5c50
to
a8a1f91
Compare
|
@bors-servo r+ thanks! |
|
📌 Commit a8a1f91 has been approved by |
2x and 4x oversampling for WaveShaperNode Issue #205 Finally managed to defeat the noise I mentioned in the issue I'm not sure if ditching FrameIterator and operating on raw vecs was a good thing 🤷♂️ Also spec mentions tail-time but I'm not sure how to handle that
|
☀️ Test successful - checks-travis |
Issue #205
Finally managed to defeat the noise I mentioned in the issue
I'm not sure if ditching FrameIterator and operating on raw vecs was a good thing 🤷♂️
Also spec mentions tail-time but I'm not sure how to handle that