Issues

• Six STT loaders ignore the cache: passed to STT.loadModel

  Blaizzy/mlx-audio-swift#279 · Filter by author magicnight opened on Oct 4, 2026
  1

• Support Confucius4-T3PO

  Blaizzy/mlx-audio-swift#278 · Filter by author hehehai opened on Sep 30, 2026
  1

• Support Confucius4-R2T2

  Blaizzy/mlx-audio-swift#277 · Filter by author hehehai opened on Sep 30, 2026
  1

• Support granite_speech5_ctc (GraniteSpeech5ForCTC) — distinct from granite_speech

  Blaizzy/mlx-audio-swift#260 · Filter by author selcukkubur opened on Sep 7, 2026
  1

• resolveOrDownloadModel's completeness check ignores additionalMatchingPatterns, so a model can load without its tokenizer

  Blaizzy/mlx-audio-swift#259 · Filter by author selcukkubur opened on Sep 7, 2026
  1

• Sortformer streaming: bounded spkcache (188-frame truncation) + no right-context loses long-range speaker identity on long audio — a memory-bounded high-accuracy operating point?

  Blaizzy/mlx-audio-swift#258 · Filter by author noorbhatia opened on Sep 5, 2026
  1

• Qwen3ASR.generate: default 20-min chunkDuration causes ~7GB fp16 KV cache and a memory-bandwidth collapse

  Blaizzy/mlx-audio-swift#249 · Filter by author ZigZagYea opened on Aug 19, 2026
  1

• Qwen3ASR.generate: maxTokens is a shared budget across chunks — long audio silently truncated

  Blaizzy/mlx-audio-swift#248 · Filter by author ZigZagYea opened on Aug 19, 2026
  1

• Mac Sequoia Support

  User request
  Blaizzy/mlx-audio-swift#243 · Filter by author linear-code opened on Aug 14, 2026
  1

• Kokoro English G2P ignores the HubCache supplied to TTS.loadModel

  Blaizzy/mlx-audio-swift#242 · Filter by author markphelps opened on Aug 8, 2026
  1

• Support Audio8-ASR

  Blaizzy/mlx-audio-swift#240 · Filter by author hehehai opened on Aug 6, 2026
  1

• Support Audio8/ARK-ASR-3B，0.6B

  Blaizzy/mlx-audio-swift#238 · Filter by author hehehai opened on Jul 31, 2026
  1

• Qwen3-TTS generation is per-step-overhead-bound, not bandwidth-bound — real-time streaming not achievable on M4 Pro with any model size

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#224 · Filter by author raucodes opened on Jul 14, 2026
  1

• Support Stable Audio 3 medium

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#210 · Filter by author JacobLinCool opened on Jun 16, 2026
  1

• Support MiMo-V2.5-ASR

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#183 · Filter by author ailuntx opened on May 12, 2026

• Support translate task for ASR models like Qwen3-ASR

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#173 · Filter by author sysulq opened on Apr 28, 2026

• Parakeet mel preprocessing bug in v0.1.2 causes ~3× WER degradation on conversational speech

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#138 · Filter by author ianandrews opened on Mar 30, 2026

• Reduce TTS model boilerplate: ~1,000 lines of duplicated loading infrastructure

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#120 · Filter by author beshkenadze opened on Mar 23, 2026
  5

• Proposal: MLXAudioKaldiAlignment

  repo/mlx-audio-swift
  Blaizzy/mlx-audio-swift#85 · Filter by author beshkenadze opened on Mar 8, 2026
  2