Issues

• HIP/ROCm build: speaker-state graph hangs in hipStreamSynchronize on audio longer than ~2 min because ARGSORT (>1024 cols) falls back to CPU; llama.cpp pin predates the upstream fix

  NVIDIA/NeMo-Speech.cpp#98 · Filter by author Warach-Kryptodian opened on Oct 31, 2026

• Windows: CUDA Backend Fails to Initialize RTX 4050 While Vulkan Backend Works

  NVIDIA/NeMo-Speech.cpp#96 · Filter by author David-H-Lau opened on Oct 31, 2026

• [Feature Request]: Add native CPU MoE offloading (--asr.salm.n_cpu_moe)

  enhancement
  NVIDIA/NeMo-Speech.cpp#95 · Filter by author BVEsun opened on Oct 30, 2026
  1

• /v1/audio/transcriptions rejects the model name the server was started with (404 unknown_model) since 0.3.0

  NVIDIA/NeMo-Speech.cpp#94 · Filter by author danilogr opened on Oct 30, 2026

• Experimental Intel NPU support through ggml-openvino: is this wanted?

  NVIDIA/NeMo-Speech.cpp#93 · Filter by author gdevenyi opened on Oct 30, 2026

• ASR language tag parser accepts <unk> as a language

  NVIDIA/NeMo-Speech.cpp#92 · Filter by author gdevenyi opened on Oct 30, 2026

• Realtime ASR with language auto: return the detected language per utterance

  NVIDIA/NeMo-Speech.cpp#90 · Filter by author gdevenyi opened on Oct 30, 2026

• serve (Vulkan): concurrent realtime sessions with ASR batching on give no output or abort with GGML_ASSERT(ne3 == ne13)

  NVIDIA/NeMo-Speech.cpp#89 · Filter by author gdevenyi opened on Oct 30, 2026

• Jetson Orin (JetPack 7.2 / CUDA 13.2): v0.2.0 cuda12 archive puts TTS on the CPU and aborts; nightly fixes that but aborts in the ggml VMM pool (cuMemAddressReserve)

  NVIDIA/NeMo-Speech.cpp#77 · Filter by author AubinGil opened on Oct 24, 2026

• ASR realtime WebSocket connections drop under load

  NVIDIA/NeMo-Speech.cpp#72 · Filter by author VitorAMJorge opened on Oct 22, 2026
  2

• Arabic streaming RNNT drops words with Q4 and FP16

  NVIDIA/NeMo-Speech.cpp#71 · Filter by author AbdeensAcademy opened on Oct 18, 2026
  1

• nemo_speech_tts_synthesize_tokens ignores options.voice_name: every pre-tokenized request is spoken by the default speaker

  NVIDIA/NeMo-Speech.cpp#64 · Filter by author scipioni opened on Oct 8, 2026

• MagpieTTS long-form: a 4–5 character sentence fails the whole request ("longform history context cache is too short")

  NVIDIA/NeMo-Speech.cpp#61 · Filter by author dxrpl opened on Oct 5, 2026

• Does this runtime support multitalker-parakeet-streaming-0.6b-v1?

  enhancement
  NVIDIA/NeMo-Speech.cpp#56 · Filter by author jfantell opened on Sep 30, 2026
  3

• Expose confidence estimation / token probabilities (currently always 1.0)

  enhancement
  NVIDIA/NeMo-Speech.cpp#55 · Filter by author NatakuZ opened on Sep 30, 2026
  2

  

• realtime WS: commit never yields .completed for utterances >= ~1s (0.5s works)

  NVIDIA/NeMo-Speech.cpp#49 · Filter by author JEFF7712 opened on Sep 20, 2026
  1

• Streaming RNNT wedges after sustained zero-PCM silence; Vulkan aborts with GGML_ASSERT(ne3 == ne13)

  NVIDIA/NeMo-Speech.cpp#48 · Filter by author BSTail opened on Sep 15, 2026

• Token-silence EOU misfires mid-sentence, hard reset corrupts transcript

  NVIDIA/NeMo-Speech.cpp#40 · Filter by author ryanleary opened on Sep 10, 2026

• Magpie TTS 2602 and 2607 for Japanese language inserts hallucinated sounds on punctuation symbols (periods).

  NVIDIA/NeMo-Speech.cpp#34 · Filter by author MidnightMurr opened on Sep 6, 2026