Opus (CELT) demo page 2: Listening tests

[xiphmont]

Greg Maxwell has just posted a nice second 'demo' page for Opus. It mostly covers the recent listening testing done by volunteers at Hydrogen Audio. Pretty colors and interactive listening/comaprison scripting!

For those of you new to Opus, it's the FOSS/RF codec we're working on within the IETF codec working group. Opus is a collaborative hybrid speech / high-fidelity audio codec built using primarily Xiph's CELT codec and the Skype SILK voice codec as inputs. That makes Opus similar in some ways to what MPEG is trying to achieve with USAC (Unified Speech and Audio Coding), though Opus is also ultra-low latency, so it looks like we're considerably ahead of MPEG here.

Comments

[(Anonymous)]

Congratulations to the CELT/Opus team! This is going to be a great codec that could surpass most existing ones.

In the short term, however, is there any plan to merge the aoTuV improvements? It was promised many time in the past. Vorbis usage now is exponentially growing due to WebM/YouTube and it would be nice to have these file compressed with the best Vorbis implementation.

Thanks!

[xiphmont.livejournal.com]

"In the short term, however, is there any plan to merge the aoTuV improvements? It was promised many time in the past."

A merge is not quite as simple as running patch. The aims of the AoTuV tuning and the reference tuning are different, and a merge also requires a code review (we can't ship anything we haven't thoroughly reviewed and tested ourselves).

It's something that's been on the weekly status and planning lists for a good long time, but other things keep bumping it back. It's a low priority exacty because it's easily available to anyone who wants it via Aoyumi :-)

[(Anonymous)]

Hi Monthy!

Will You (i.e. Xiph...) try to take the samples wich failed in the test (01,02,19,26) and try to further tune the CELT/Opus? Or would it break other samples' quality?