Uzbek automatic speech recognition, fine-tuned from openai/whisper-small on 753 hours of exclusively human-transcribed Uzbek speech.
This is the small (244M parameters) model in the NavAI Whisper family. It is trained only on data with genuine human transcripts — no pseudo-labels, no scraped audio — and no evaluation test set is present in the training data.
Model overview
The NavAI Whisper family spans four sizes so you can trade accuracy for latency:
NavAI Uzbek Whisper family — macro-WER vs model size
Macro WER is the mean Word Error Rate across four held-out Uzbek test sets (see Evaluation). Lower is better. All four models live in the NavAI Whisper Collection.
Passing a file path straight to asr("audio.wav", ...) also works, but the pipeline decodes it with ffmpeg, which must be installed separately (conda install -c conda-forge ffmpeg). Loading the array with librosa as above needs no ffmpeg.
Batched transcription:
python
1files =["a.wav","b.wav","c.wav"]2batch =[{"array": librosa.load(f, sr=16000)[0],"sampling_rate":16000}for f in files]3for out in asr(batch, batch_size=8, generate_kwargs={"language":"uz"}):4print(out["text"])
Best practices
Language token. Pass generate_kwargs={"language": "uz"} for best results.
Long audio. Set chunk_length_s=30; Whisper transcribes in 30-second windows.
Repetition guard. The shipped generation_config sets no_repeat_ngram_size=4, and all reported numbers are scored with it on.
Script. Output is Latin-script Uzbek (uzn_Latn).
Model data
Training dataset
753 hours of human-transcribed Uzbek speech, from four openly-licensed corpora:
crowd-sourced read speech; distributed on Hugging Face as DavronSherbaev/uzbekvoice-filtered (Apache-2.0), linked from the official site (accessed 19 July 2026)
Explicitly excluded from training: machine pseudo-labels, scraped YouTube audio, and every evaluation test set (including FeruzaSpeech and FLEURS test). No evaluation data appears in training.
Attribution (USC, CC BY 4.0)."Our product uses ISSAI and TUIT Uzbek Speech Corpus, which is available under a Creative Commons Attribution 4.0 International License." — Musaev et al., USC: An Open-Source Uzbek Speech Corpus (arXiv:2107.14419).
Data preprocessing
Transcripts normalized with the released uzbek_text_norm package: Cyrillic→Latin folding, spelled-out numbers/ordinals, Latin apostrophe folding (oʻ/gʻ), and punctuation handling.
Audio resampled to 16 kHz mono; log-mel features padded to 30 seconds.
Training procedure
Full fine-tune of openai/whisper-small on the 753 h corpus above. Hyperparameters were selected per size by a hyperparameter search:
This size (search winner): HPO winner (lr 2.16e-5, wd 0.05, mtp 0.03)
Shared recipe: effective batch size 64 · AdamW · linear warmup (10%) → linear decay · up to 10 epochs · SpecAugment on the log-mel features · mixed precision (bf16 on A100 / fp16 on V100) · 4×GPU data-parallel (DDP) · repetition guard no_repeat_ngram_size=4.
Model selection: hyperparameters and the released checkpoint were selected on validation WER over a held-out USC split; the reported figures are then measured on the four-domain benchmark (FLEURS · Common Voice · USC · FeruzaSpeech).
Evaluation
Benchmark sets
Four held-out Uzbek test splits, spanning distinct acoustic and lexical conditions:
Decoding: greedy for every system — num_beams=1, no sampling. No model is advantaged by beam search.
Normalizer: the released uzbek_text_norm (v0.3.0, Apache-2.0), applied symmetrically to reference and hypothesis for every system: lowercase, all punctuation removed, hyphens → space (splitting hyphenated compounds, matching Whisper's BasicTextNormalizer), every apostrophe variant folded to the Uzbek okina (oʻ/gʻ), Cyrillic→Latin folding, and digits/ordinals spelled out. pip install uzbek-text-norm to reproduce every number in this card.
Uniform re-decode: every system — ours and baselines — is re-decoded on one harness with the repetition guard on (no_repeat_ngram_size=4), so none is advantaged by its own decode defaults.
Common Voice references were patched (2026-07-12) to remove a metadata bug that corrupted two references and inflated CV WER for every system. Any CV number published before that date — including in earlier revisions of this card — is unreliable.
Macro WER: mean of per-set WER (each domain weighted equally).
Limitations
Encyclopedic / named-entity speech is the weakest condition — rare foreign proper nouns are missed more often than common vocabulary.
Audio is processed in 30-second windows; extremely long files are chunked automatically.
Trained for Latin-script Uzbek; Cyrillic input is normalized at scoring but is not a primary target.
Far-field, noisy, or heavy-dialect audio is less well covered than clean read/conversational speech.
Ethical considerations
This model transcribes Uzbek speech and inherits biases present in its training data (speaker demographics, recording conditions, vocabulary). It should not be used as the sole basis for consequential decisions about individuals. Transcriptions of sensitive audio should be handled under appropriate privacy safeguards.
License
Released under the Apache License 2.0. Model weights are licensed independently of the training data; each training source retains its own license (documented above).
Citation
If you use these models or the uzbek_text_norm benchmark, please cite:
bibtex
1@misc{navai_whisper_uzbek_2026,
2 title = {NavAI Whisper: Open Uzbek Speech Recognition},
3 author = {NavAI},
4 year = {2026},
5 howpublished = {\url{https://huggingface.co/navai-uz}}
6}
Data sources
bibtex
1@inproceedings{musaev2021usc,
2 title = {USC: An Open-Source Uzbek Speech Corpus and Initial Speech Recognition Experiments},
3 author = {Musaev, Muhammadjon and others},
4 booktitle = {Speech and Computer (SPECOM)},
5 year = {2021},
6 note = {arXiv:2107.14419}
7}
8@misc{commonvoice22,
9 title = {Common Voice Corpus 22.0 (Uzbek)},
10 author = {{Mozilla Foundation}},
11 url = {https://commonvoice.mozilla.org/en}
12}
13@article{conneau2022fleurs,
14 title = {FLEURS: Few-shot Learning Evaluation of Universal Representations of Speech},
15 author = {Conneau, Alexis and others},
16 journal = {arXiv:2205.12446},
17 year = {2022}
18}
