Automatic Speech Recognition
NeMo
ONNX
Safetensors
GGUF
parakeet_tdt
parakeet
tdt
sherpa-onnx
multilingual
speech-recognition
gabor
fastconformer
Eval Results (legacy)
Instructions to use oruk/orukeet with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use oruk/orukeet with NeMo:
import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.ASRModel.from_pretrained("oruk/orukeet") transcriptions = asr_model.transcribe(["file.wav"]) - Notebooks
- Google Colab
- Kaggle
Commit History
3 days ago
-
Link runnable hosted API recipes and clarify training partition wording
Verified signatureDocumentation only: distinguish hosted English preview, subscription allowance and optional task pricing from local multilingual weights; link public examples. Preserve all benchmark measurements and numeric training counts.d16ded2 -
Add the Hosted API section (English, 60 s, 4 MiB, no word timestamps, plan allowance)
Verified signatureDiscloses the hosted preview's limits from the oruk.ai OpenAPI contract and what hosted adds over the local weights (no install or GPU, optional emotion scores and speaker diarization). Links carry utm_source=huggingface&utm_medium=model-card&utm_campaign=orukeet. Everything else on the card is unchanged.ffa186b -
Add crediting block, oruk.ai links, questions pointer and FLEURS model-index
Verified signatureREADME: a 'Crediting Orukeet' copy-paste block (Orukeet by Oruk, built on Parakeet-TDT v3; weights on Hugging Face; CC BY-SA 4.0; https://oruk.ai/models/orukeet), oruk.ai links tagged utm_source=huggingface&utm_medium=model-card&utm_campaign=orukeet, a 'Questions and bugs' pointer (GitHub issues for bugs, Discussions for model questions; no language-forcing control; hosted Orukeet is English-only), and the license sentence 'Code is MIT; weights and fitted kernels are CC BY-SA 4.0' linking LICENSE and LICENSE-WEIGHTS. Front matter: a model-index with FLEURS only (pooled 25-language 9.85% and English 3.82% WER from the card's own table); LibriSpeech test-other is excluded because it was used for adaptation. No hosted-API section: the origin probe timed out and /readyz does not list Orukeet. From the 2026-11-10 STT audit, section 8 (F047, F157, F147, G2604, G2702, G1301, G1308). Mirrors Oruk-AI/orukeet PR #12.de5d8c3
-
Point integration guide at qualified documentation revision
Verified signatureb59c13a -
Publish bounded matched-PCM native Slovenian diagnosis
Verified signature7df0b47 -
Publish bounded OpenWhispr consumer parity and warmup qualification
Verified signaturefe3bbba -
Document full 25-language Core ML qualification and current GPT comparison
Verified signatureb3421ca -
Publish opt-in iOS INT8 integration metadata and device qualification scope
Verified signature76f5e73 -
Stage Orukeet INT8 Core ML candidate for OpenWhispr draft integration (#2)
- Stage Orukeet INT8 Core ML candidate for OpenWhispr draft integration (419d7f79e290127e202a0f610509868d314743eb)eea14df
-
Remove retired benchmark material
Verified signaturedebfb0d
-
Withdraw Open ASR submission; preserve research artifacts
Verified signature14ab744
-
Publish audited public-eight Open ASR H200 results and training disclosure
Verified signatureef0ccd4
-
Add qualified Transformers FP32 export for standard Parakeet clients
Verified signatureStrict mapping of all723 source tensors; reproducible conversion; actual Buzz offline/multilingual checks;120-clip export comparison reports5.43% vs source-NeMo5.34% WER separately from published model benchmark. Existing artifacts and metadata preserved.10b603f -
Add verified standalone sherpa ONNX files for native model catalogs
Verified signatureeac739d
-
Publish verified Orukeet r3 Core ML preview bundles for TapTalk
Verified signature43142dd
-
Update README.md
Verified signaturee31d65f
-
Add validated combined INT8 ONNX export for local Rust engines
Verified signature1751fce
-
Expose transcribe.cpp Q8 at a Handy-compatible root filename
Verified signaturebc4e7f1 -
Add validated transcribe.cpp Q8 export of Orukeet r3
Verified signatureb5f8f87 -
Link verified OpenWhispr file-upload steps
Verified signaturee0c97b3
-
Link the recorded local transcription walkthrough
Verified signatureeca8ab0 -
Update native quickstart and link verified deployment options
Verified signature6f4afeb
-
Update oruk branding to the current Signal tile and wordmark
Verified signaturec8b2516 -
Document the faster ONNX export and paired OpenWhispr validation
Verified signatureb34b477 -
Speed up the sherpa ONNX export with exact depthwise arithmetic
Verified signature55a984d -
Document the shared sherpa-onnx deployment and OpenWhispr validation
Verified signature2dc1ce8 -
Add validated Parakeet TDT v3 ONNX INT8 deployment format
Verified signature74673cf
-
Link the public OpenWhispr integration PR
Verified signaturedf78b75 -
Prepare the Orukeet v0.1.0 model card, report, benchmark scores and citations
Verified signature769e303 -
Add Orukeet v0.1.0 inference package and example
Verified signature97f7de5 -
Add the final Orukeet v0.1.0 source, Q8 and F16 weights
Verified signature555136b -
initial commit
Verified signature6e1b766