Skip to main content

Minutes IETF124: mlcodec: Thu 22:00
minutes-124-mlcodec-202511062200-00

Meeting Minutes Machine Learning for Audio Coding (mlcodec) WG
Date and time 2025-11-06 22:00
Title Minutes IETF124: mlcodec: Thu 22:00
State Active
Other versions markdown
Last updated 2025-11-17

minutes-124-mlcodec-202511062200-00

Machine Learning for Audio Coding (mlcodec) Working Group
MLCODEC @ IETF 124 in Montreal, Canada
17:00-19:00 UTC-5 Thursday, November 6, 2025, room Notre Dame
Chairs: Greg Maxwell, Mo Zanaty
Area Director: Orie Steele
Notetaker: Jonathan Lennox

Administrivia - Chairs, 5 min

Note well, new area director, agenda bash, draft status

Agenda bash: speech coding enhancements moved before deep redundancy

Opus extension mechanism - Timothy Terriberry, 10 min

draft-ietf-mlcodec-opus-extension

Post-WGLC version published

Examples added, reviews of that, and draft in general, would be helpful.
Will do second WGLC.

Jonathan Lennox: Multi-channel opus proposal in AVTCore should copy the
SDP mechanism from this draft.

Jean-Marc Valin: We should say that this document updates both the Opus
draft and the Opus RTP draft

Mail to be sent (by Mo?) to AVT mailing list to mention this.

Speech coding enhancements - Jan Buethe, 20 min

draft-ietf-mlcodec-opus-speech-coding-enhancement

Will continue iterating, draft update, work out where to host test data.

Scalable quality extension - Jean-Marc Valin, 20 min

draft-ietf-mlcodec-opus-scalable-quality-extension

Format is stable, no further changes planned

Jonathan Lennox: this will need SDP parameters

Deep REDundancy - Jean-Marc Valin, 20 min

draft-ietf-mlcodec-opus-dred

Probable final model chosen, please test, and review draft
Need to work out where to publish large artifacts - Orie Steele (AD) to
talk to IANA

Speech quality and intelligibility testing - Laura Lechler, 20 min

draft-lechler-mlcodec-test-battery

Generally good results for final models

Orie: The methodology should live on, even if the specific results
aren't published

Jean-Marc: specific results are useful for user of DRED to know what
they can expect in intelligibility and quality

Orie says charter can be updated for publishing informational documents.

Mo: suggest putting DRED results along with other DRED artifacts.

Chairs: should we adopt methodology as a work item of this group?

Poll: 17 yes, 0 no, 1 no opinion

Laura to submit as WG document

AOB

Will you attend in person in Shenzen?

5 yes, 7 no