All issues

Search results

PocketTTS: ensureModels() downloads unused precision/placement variants, and file downloads run sequentially

Status: Open.

#853 In FluidInference/FluidAudio;

· twillixa opened on Aug 12, 2026

KokoroTail iSTFT: the COLA-corrected tail from #699 is not in the published bundles — ANE-ja/ is byte-identical to ANE/, and en/zh still need the fix

Status: Open.

#852 In FluidInference/FluidAudio;

· drakulavich opened on Aug 12, 2026

Vocabulary boosting / CTC keyword spotting on the unified Parakeet path?

Status: Open.

#851 In FluidInference/FluidAudio;

· darylthornhill opened on Aug 12, 2026

Parakeet TDT v3 batch: mixed-language window deletes the entire English opening (deterministic; truncated audio decodes perfectly)

Status: Open.

#850 In FluidInference/FluidAudio;

· nachoal opened on Aug 11, 2026

160 ms Nemotron streaming tier: the checkpoint's native [70,1] lookahead — conversion validated on your harness, plumbing patch ready

Status: Open.

#846 In FluidInference/FluidAudio;

· nischith-memfold opened on Aug 11, 2026

Custom vocabulary for Nemotron streaming: decode-time RNN-T biasing (no CTC head required)

Status: Open.

#841 In FluidInference/FluidAudio;

· nischith-memfold opened on Aug 8, 2026

applyEnglishBlocklist corrupts clean German/Dutch/Italian/Polish: the French-tuned token blocklist runs for all non-English Latin languages

Status: Open.

#840 In FluidInference/FluidAudio;

· saurabhav88 opened on Aug 8, 2026

TextNormalizer is a silent no-op for SwiftPM consumers: no native ITN library ships, dlopen(nil) finds nothing, normalize() returns input unchanged (0.15.5)

Status: Open.

#839 In FluidInference/FluidAudio;

· rico-suave-23 opened on Aug 7, 2026

nemotron-multilingual 560ms streaming: isolated short word silently decodes to blank, controlled by the preceding audio

Status: Open.

#838 In FluidInference/FluidAudio;

· nischith-memfold opened on Aug 7, 2026

KokoroAne: vocoder fails on macOS 14 under every compute-unit preset

Status: Open.

#836 In FluidInference/FluidAudio;

· drakulavich opened on Aug 5, 2026

ASR long-form: mid-file window ending inside a dead-silence run drops nearby speech (same pathology as #747, interior windows)

bugSomething isn't workingSomething isn't working

Status: Open.

#803 In FluidInference/FluidAudio;

· Alex-Wengg opened on Jul 16, 2026

Offline diarizer: auto clustering and numSpeakers=2 return the same speaker count but different partitions (69% vs 92% correctly-attributed speech)

Status: Open.

#801 In FluidInference/FluidAudio;

· kasuki511 opened on Jul 16, 2026