# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("WindstormLabs/listen-windy-lingua-he", dtype="auto")WindyWord.ai STT — Hebrew Lingua (GPU (safetensors))
Transcribes Hebrew speech (Afro-Asiatic > Semitic).
Note: Replaces a previous build whose weights were incomplete (decoder layers 10-23 missing) and produced gibberish output. Now derived from
oridror/whisper-large-v3-turbo-hebrew-r1-myd-r1(Whisper Large-v3 turbo Hebrew fine-tune). Verified post-upload at WER 24.2% / CER 11.5% / script-match 99% on 20-sample FLEURS he_il — GOOD tier. Tokenizer/preprocessor files filled in fromopenai/whisper-large-v3since the upstream fine-tune omits them.
Quality
- FLEURS WER: 66.9% (50-sample audit)
- CER: 0.3902
- Tier: UNUSABLE-GAP ⭐
- Source: WindyWord Grand Rounds v2 audit (50-sample FLEURS)
About this variant
This is the safetensors deployment format of our Hebrew Lingua STT model. Load it via the safetensors/ subfolder.
Part of the WindyWord.ai STT fleet — covering 35+ languages that commercial speech-to-text APIs underserve, with proper dialect / script disclosures where they matter.
Usage
from transformers import WhisperForConditionalGeneration, WhisperProcessor
processor = WhisperProcessor.from_pretrained("WindyWord/listen-windy-lingua-he", subfolder="safetensors")
model = WhisperForConditionalGeneration.from_pretrained("WindyWord/listen-windy-lingua-he", subfolder="safetensors")
Commercial Use
Visit windyword.ai for apps and API access.
Provenance & License
Weights derived from upstream community Whisper fine-tunes (see individual model card for exact lineage). Redistributed under Apache-2.0 (inherited).
Certified by Opus 4.6 Opus-Claw (Dr. C) on Veron-1 (RTX 5090, Mt Pleasant SC).
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="WindstormLabs/listen-windy-lingua-he")