Topics

Table of Contents

  1. Low-resource low-bitrate speech coding methods
  2. Broader impacts and applications
  3. Evaluation, benchmarking, and reproducibility

Topics areas for the 2026 LRAC Workshop center around low-resource, efficient neural speech codecs and include, but are not limited to:

Low-resource low-bitrate speech coding methods

  • Solution papers for the 2025 LRAC Challenge entries

  • Neural speech coding under joint compute, bitrate, and latency constraints

  • Quantization techniques for efficient speech codecs

  • Model complexity and size reduction strategies

  • Generative machine learning approaches for speech codecs

  • Joint speech enhancement and coding for practical applications

  • Designing for robustness to noise and reverberation

  • Efficient modeling of semantic, acoustic, and prosodic information in speech codecs

  • Resilience to network conditions for streaming low-resource speech codecs

Broader impacts and applications

  • Neural speech codecs as universal feature extractors for downstream tasks, e.g., spoken dialogue systems, multimodal language models

  • Open challenges for real-world deployment of neural speech codecs

  • Application domains (e.g., IoT, edge) for low-resource speech codecs

Evaluation, benchmarking, and reproducibility

  • Objective metrics and listening test methodologies for evaluation

  • Availability and use of public train/test datasets

  • Data curation for speech codec development

  • Ensuring reproducibility in codec research

Back to top