NVIDIA-NeMo/Speech

View on GitHub

A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)

Stars: 18.5k
Forks: 3.6k
Open beginner issues: 0
Indexed issues: 137
Avg merge: 3d 21h
Merged PRs (30d): 60

Dominant language: Python
License: Apache-2.0
Last GitHub push: Sep 18, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: help wanted

Beginner friendly Everything

Open All

137 open issues indexed

nemotron-labs-voicechat-11b: Training data Open

community-request

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#16286 · Opened Sep 18, 2026
Sortformer splits one speaker into two when their distance to the mic changes Open

community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 30/100

NVIDIA-NeMo/Speech#16278 · Opened Sep 17, 2026
nemotron-labs-voicechat: Issues with the model's voice Open

bug community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#16252 · Opened Sep 14, 2026
MagpieTTS raw-audio padding adds a spurious silent frame to exact-multiple-length audio Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 84/100

NVIDIA-NeMo/Speech#16249 · Opened Sep 13, 2026
Early-interruption augmentation can drop the relocated EOS token entirely Open

community-request

Difficulty 2/5 1-3 hours Newbie friendliness 86/100

NVIDIA-NeMo/Speech#16247 · Opened Sep 12, 2026
SubsamplingReductionModule pooling returns wrong lengths (factor 4 -> [6,5], factor 8 -> [0,0]) Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

NVIDIA-NeMo/Speech#16224 · Opened Sep 7, 2026
ConvSubsampling forward broken: chunking=-1, striding_conv1d/dw_striding_conv1d, and vggnet all raise TypeError Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 68/100

NVIDIA-NeMo/Speech#16223 · Opened Sep 7, 2026
SimpleSegmentedTextAggregator never treats a sentence ending in a decimal number as complete Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

NVIDIA-NeMo/Speech#16218 · Opened Sep 5, 2026
FastPitchModel_SSL / SSLDisentangler crash on validation: on_validation_epoch_end() missing 1 required positional argument: 'outputs' Open

community-request

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

NVIDIA-NeMo/Speech#16216 · Opened Sep 5, 2026
[Bug] NeMo Docker image works on single GPU but not on multi-GPU setup Open

bug community-request waiting-on-maintainers

Difficulty 4/5 3-5 days Newbie friendliness 32/100

NVIDIA-NeMo/Speech#16211 · Opened Sep 3, 2026
[ALERT] Access to NVIDIA's Self-Hosted Runners is Expiring Open

nvidia-runners

Difficulty 2/5 1-3 hours Newbie friendliness 35/100

NVIDIA-NeMo/Speech#16207 · Opened Sep 2, 2026
IPLEpochStopper ignores enable_stop, always requests a training stop Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

NVIDIA-NeMo/Speech#16182 · Opened Sep 1, 2026
AudioAugmenter loudness-norm disablement leaks across batch rows in add_room_ir_to_batch / add_mic_ir_to_batch Open

community-request

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

NVIDIA-NeMo/Speech#16180 · Opened Sep 1, 2026
voice_agent: NemoSTTService's default model is misspelled ('nnvidia'), silently defaulting has_turn_taking to False Open

community-request waiting-on-maintainers

Difficulty 1/5 Under an hour Newbie friendliness 92/100

NVIDIA-NeMo/Speech#16178 · Opened Sep 1, 2026
verify_speakers_batch returns an unindexable 0-d array for a single audio pair Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

NVIDIA-NeMo/Speech#16172 · Opened Aug 30, 2026
[TTS] AlignerModel bin_loss warmup ramp (bin_loss_scale) has no effect Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 85/100

NVIDIA-NeMo/Speech#16170 · Opened Aug 30, 2026
prepare_text_and_asr_labels silently drops EOS when delay_text_channel_by shifts it into the delay window Open

community-request

Difficulty 2/5 1-3 hours Newbie friendliness 86/100

NVIDIA-NeMo/Speech#16164 · Opened Aug 29, 2026
malsd_batch beam search decodes labels for a batch row with encoder_output_length == 0 Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 74/100

NVIDIA-NeMo/Speech#16161 · Opened Aug 28, 2026
Verify evals on Papers with Code Open

community-request waiting-on-customer

NVIDIA-NeMo/Speech#16156 · 1 comment · 1 assignee · Opened Aug 28, 2026
Abnormal Punctuation and Excessive Sentence Segmentation in German (canary-180m-flash) Open

bug community-request waiting-on-maintainers

Difficulty 4/5 3-5 days Newbie friendliness 38/100

NVIDIA-NeMo/Speech#16154 · 3 comments · Opened Aug 27, 2026
EncDecHybridRNNTCTCBPEEOUModel.validation_pass raises UnboundLocalError: text_pred on its default path Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

NVIDIA-NeMo/Speech#16150 · Opened Aug 27, 2026
Relax the `lightning<=2.4.0` constraint for NeMo Speech ASR Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

NVIDIA-NeMo/Speech#16144 · Opened Aug 26, 2026
[Nemotron VoiceChat] ASR EOS loss-weight mismatch between labels and token weighting Open

bug community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

NVIDIA-NeMo/Speech#16142 · Opened Aug 26, 2026
Enable FP16/BF16 compute_dtype for cache-aware streaming ASR inference Open

bug community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 72/100

NVIDIA-NeMo/Speech#16137 · Opened Aug 25, 2026
Question: text-token bottleneck vs direct S2S in NemotronLabs VoiceChat Open

community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#16136 · Opened Aug 25, 2026
S2S model with multi stage train Open

community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#16134 · Opened Aug 25, 2026
Nemotron Labs VoiceChat image for streaming interaction Open

community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#16118 · Opened Aug 20, 2026
UnboundLocalError: storage_options when an EMA callback is present and save_last_n_optim_states >= 0 Open

community-request

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

NVIDIA-NeMo/Speech#16111 · Opened Aug 18, 2026
exp_manager S3 resume check runs on every rank because the guard ANDs two mutually exclusive URL schemes Open

community-request

Difficulty 1/5 Under an hour Newbie friendliness 88/100

NVIDIA-NeMo/Speech#16110 · Opened Aug 18, 2026
ENCharParser(abbreviation_version='expanded') returns None for every input Open

community-request

Difficulty 2/5 1-3 hours Newbie friendliness 86/100

NVIDIA-NeMo/Speech#16108 · Opened Aug 18, 2026
FeatureLabel.uniq_labels collects the characters of each label, breaking FeatureToLabelDataset Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

NVIDIA-NeMo/Speech#16106 · Opened Aug 18, 2026
random_segment augmentation raises TypeError: pad_width must be of integral type whenever it has to pad Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 78/100

NVIDIA-NeMo/Speech#16104 · Opened Aug 18, 2026
save_adapters/load_adapters do not round-trip: subset saves fail to load, and names are matched by substring Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 48/100

NVIDIA-NeMo/Speech#16102 · Opened Aug 18, 2026
Lhotse RIR augmentation is seeded from seed instead of shard_seed, so all ranks draw identical reverberation Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

NVIDIA-NeMo/Speech#16100 · Opened Aug 18, 2026
Frame offset is 10x too large when merging hour-long chunk hypotheses Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 35/100

NVIDIA-NeMo/Speech#16099 · Opened Aug 18, 2026
Corpus average_cpWER still uses the concatenation algorithm that PR #15573 replaced Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 76/100

NVIDIA-NeMo/Speech#16093 · Opened Aug 18, 2026
Docs name the multitask metrics config key multitask_metrics_config; the code reads multitask_metrics_cfg Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

NVIDIA-NeMo/Speech#16091 · Opened Aug 18, 2026
Shipped SALM and beamforming configs name _target_ classes and modules that do not exist Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

NVIDIA-NeMo/Speech#16089 · Opened Aug 18, 2026
Phoneme dictionary entries above U+00FF are dropped silently; ACCENTED_CHARS excludes 51 of 93 declared Vietnamese graphemes Open

community-request waiting-on-maintainers

Difficulty 4/5 3-5 days Newbie friendliness 38/100

NVIDIA-NeMo/Speech#16083 · Opened Aug 18, 2026
max_open_streams with a multi-manifest config raises AttributeError: LazyNeMoIterator has no attribute to_shards Open

community-request waiting-on-maintainers

Difficulty 3/5 1-2 days Newbie friendliness 68/100

NVIDIA-NeMo/Speech#16081 · Opened Aug 18, 2026
DALI dataset silently accepts invalid window names: validation is unreachable dead code (dict.get in try/except) Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 45/100

NVIDIA-NeMo/Speech#16073 · 1 comment · Opened Aug 18, 2026
Chunked long-audio timestamps drift monotonically (2.36 s over 1 hour) because chunk offsets are floor-divided before accumulation Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

NVIDIA-NeMo/Speech#16072 · Opened Aug 18, 2026
[Feature] Support Hugging Face Storage Buckets for checkpointing Open

community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 30/100

NVIDIA-NeMo/Speech#16016 · Opened Aug 3, 2026
Lhotse error when trainer.limit_train_batches=1.0 Open

bug community-request

Difficulty 3/5 1-2 days Newbie friendliness 68/100

NVIDIA-NeMo/Speech#16004 · Opened Jul 31, 2026
[BUG] fuse_loss_wer=true deadlocks multi-GPU validation when validation_ds.batch_size > fused_batch_size Open

bug community-request

Difficulty 4/5 3-5 days Newbie friendliness 55/100

NVIDIA-NeMo/Speech#16003 · 2 comments · Opened Jul 31, 2026
[audio.cpp] CUDA-optimized C++ implementation of Nemotron-3.5-ASR-Streaming-0.6B. 400.0x realtime on RTX 5090 Open

community-request waiting-on-maintainers

Difficulty 2/5 1-3 hours Newbie friendliness 42/100

NVIDIA-NeMo/Speech#15971 · Opened Jul 27, 2026
canary-1b-v2 systematically drops word-final Greek sigma «ς» (U+03C2) — the final-sigma character is absent from the SentencePiece tokenizer vocabulary Open

community-request waiting-on-customer

Difficulty 5/5 Over a week Newbie friendliness 35/100

NVIDIA-NeMo/Speech#15936 · 1 comment · Opened Jul 23, 2026
Nemotron-3-Ultra-free: Behavioral Alignment Failure — Distrust-by-Default, Constraint Violations, Zero Intra-Session Learning Open

bug community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#15922 · 1 reaction · Opened Jul 18, 2026
Speaker Diarization to parakeet v3 and nemotron-3.5-asr-streaming-0.6b Open

community-request waiting-on-maintainers

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA-NeMo/Speech#15919 · Opened Jul 17, 2026
No Streaming API for streaming sortformer? Open

community-request waiting-on-maintainers

Difficulty 4/5 3-5 days Newbie friendliness 45/100

NVIDIA-NeMo/Speech#15918 · 4 comments · Opened Jul 17, 2026

NVIDIA-NeMo/Speech

Get new issues in your inbox