whisper-ja-760M

For usage instructions follow openai/whisper-large-v3-turbo. Note for faster-whisper vocab changes make model.ismultilingual and suppresstokens wrong. Please adjust the code as required if you want to use this with faster-whisper. Turbo finetune with japanese tokenizer. Full finetune trained 2^19 steps, batch size 64. Smaller vocab with ~1.6x bytes/token allows faster speed with 4 layers vs 2 layer distil (10% larger decoder). Benchmarks. Short form slightly behind v0.2 (trained less?) but long form much better. Also trained for lyrics but untested. Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC)

Model · Speech recognition

whisper-large-v3-turbo

OpenAI

Whisper is a state-of-the-art model for automatic speech recognition (ASR) and speech translation, proposed in the paper Robust Speech Recognition via Large-Scale Weak Supervision by Alec Radford et al. from OpenAI. Trained on >5M hours of labeled data, Whisper demonstrates a strong ability to generalise to many datasets and domains in a zero-shot setting. Whisper large-v3-turbo is a finetuned version of a pruned Whisper large-v3. In other words, it's the exact same model, except that the number of decoding layers have reduced from 32 to 4. As a result, the model is way faster, at the expense of a minor quality degradation. You can find more details about it in this GitHub discussion.…

Fine-tuned openai/whisper-large-v3-turbo on Monsoon Hindi ASR data using LoRA (r=32, alpha=64). English capabilities preserved with <10% relative degradation across all benchmarks. LoRA's structural separation (frozen base + low-rank adapters) provides a natural safeguard against catastrophic forgetting.

Model · Speech recognition

nemotron-3.5-asr-streaming-0.6b

NVIDIA

/ Improve list spacing / / Badge alignment consistency / Nemotron 3.5 ASR is a multilingual, streaming Automatic Speech Recognition (ASR) model engineered to deliver high-quality multilingual transcription across both low-latency streaming and high-throughput batch workloads. Developed by NVIDIA, this 600M parameter model transcribes speech into text with native support for punctuation and capitalization, and offers runtime flexibility with configurable chunk sizes, including 80ms, 160ms, 320ms, 560ms, and 1120ms. By leveraging a state-of-the-art Cache-Aware FastConformer-RNNT architecture, the model eliminates redundant overlapping computations common in traditional "buffered" streaming.…

An equal-weight average of the epoch 6, 7, 8, 9 and 10 checkpoints of No training was done here, and inference costs exactly what one model costs. Like its ingredient, this model has no reported WER or CER and cannot have one. That run trains on every held-out hour the project has, including the SAPC2 dev split the rest of this family scores against. Averaging its epochs does not create a set to measure on. So this checkpoint rests on a bet rather than a measurement, and it is worth same soup of the same five epochs was worth 0.27 CER points — 6.06% against 6.33% for the best single epoch. That is the whole of the evidence. It is evidence from a different architecture (RNN-T, not TDT) on…

nvidia/parakeet-tdt-0.6b-v3 fine-tuned on all 1,047.7 hours this project holds: SAPC1 train and dev, SAPC2 train, the SAPC2 dev split the rest of this family scores against, 103.1 hours of synthetic dysarthric speech, 79.2 hours recovered by force-aligning and cutting recordings past the 45-second training cap, and 15.5 hours of AtaxiaUK and HeyJay!, which are outside the challenge corpora and make this an unconstrained-track model. This model has no reported WER or CER, and cannot have one. Every held-out hour is in its training data. That was the point: the hyperparameters were settled on the sibling runs that do hold out a dev split, and this run spends that split as training data…