Issues

• Cohere Transcribe: default computeUnits .all puts the int8 encoder on the GPU, footprint 2.1 GB → 12.4 GB after the first call; .cpuAndNeuralEngine stays ~80 MB

  FluidInference/FluidAudio#1045 · Filter by author daghreri opened on Oct 19, 2026

• KokoroAne G2P: "g2p-only" download pulls every root-level .json (~30 MB of es/fr/gb/us gold-silver lexicons unused by English)

  FluidInference/FluidAudio#1044 · Filter by author bhavik-mangla opened on Oct 17, 2026

• KokoroAne v3: ensure() re-checks and re-downloads .mlpackage sources the compiled .mlmodelc replaces (~190 MB kept on disk, hashed every launch)

  FluidInference/FluidAudio#1043 · Filter by author bhavik-mangla opened on Oct 17, 2026

• Long-form Japanese (tdtJa) silently drops text at chunk seams: the splice guard assumes spaced words

  FluidInference/FluidAudio#1031 · Filter by author ertra opened on Oct 15, 2026

• Encoder.mlmodelc (parakeet-ultra) takes ~105 s to load on iPhone 18 Pro Max vs. ~12 s on iPhone 16 Pro Max after each app update

  FluidInference/FluidAudio#1022 · Filter by author thomas-hensel opened on Oct 13, 2026
  2

• Full-duplex recognition research

  FluidInference/FluidAudio#1007 · Filter by author BenRacicot opened on Oct 10, 2026
  2

• CampPlusEmbedder: embeddings differ from reference CAM++ (fbank not mean-normalized, Hamming window, pooling divisor)

  FluidInference/FluidAudio#985 · Filter by author SeoFood opened on Oct 4, 2026

• Batch Parakeet v3 custom vocabulary: measurements on 500 real dictations and a real-speech exam, with audio (great recall, common words replaced, a few bugs)

  FluidInference/FluidAudio#967 · Filter by author saurabhav88 opened on Sep 26, 2026
  1

• v0.17.0: window merge duplicates words on long audio (single-window decode of the same audio is correct)

  FluidInference/FluidAudio#954 · Filter by author maboa opened on Sep 24, 2026
  6

• PocketTTS: a sentence that must be cut is cut between words although a dash is available

  FluidInference/FluidAudio#934 · Filter by author arvindvenkataramani opened on Sep 19, 2026

• PocketTTS: text containing a parenthesis is garbled, and the first sentence after a paragraph break is cut mid-phrase

  FluidInference/FluidAudio#931 · Filter by author arvindvenkataramani opened on Sep 19, 2026

• SlidingWindowAsrManager: spurious text injected at chunk boundary when minContextForConfirmation/confirmationThreshold are zeroed

  FluidInference/FluidAudio#912 · Filter by author janisbelozerovs-dev opened on Sep 13, 2026

• SlidingWindow streaming: default rescue floors replace an unrelated opening word with a vocabulary term ("Hey" → "Codex")

  FluidInference/FluidAudio#899 · Filter by author Alex-Wengg opened on Sep 9, 2026
  2

• Model Support Requests

  enhancementquestion
  Documentation

  FluidInference/FluidAudio#49 · Filter by author BrandonWeng opened on Jul 30, 2025
  89