Flavors of Moonshine: Tiny Specialized ASR Models for Edge Devices

• URL: http://arxiv.org/abs/2509.02523v1
• Date: Tue, 02 Sep 2025 17:22:54 GMT
• Title: Flavors of Moonshine: Tiny Specialized ASR Models for Edge Devices
• Authors: Evan King, Adam Sabra, Manjunath Kudlur, James Wang, Pete Warden,
• Abstract summary: We present a suite of tiny automatic speech recognition (ASR) models specialized for a range of underrepresented languages.<n>We release Arabic, Chinese, Japanese, Korean, Ukrainian, and Vietnamese models under a permissive open-source license.
• Score: 1.4625828590961276
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We present the Flavors of Moonshine, a suite of tiny automatic speech recognition (ASR) models specialized for a range of underrepresented languages. Prevailing wisdom suggests that multilingual ASR models outperform monolingual counterparts by exploiting cross-lingual phonetic similarities. We challenge this assumption, showing that for sufficiently small models (27M parameters), training monolingual systems on a carefully balanced mix of high-quality human-labeled, pseudo-labeled, and synthetic data yields substantially superior performance. On average, our models achieve error rates 48% lower than the comparably sized Whisper Tiny model, outperform the 9x larger Whisper Small model, and in most cases match or outperform the 28x larger Whisper Medium model. These results advance the state of the art for models of this size, enabling accurate on-device ASR for languages that previously had limited support. We release Arabic, Chinese, Japanese, Korean, Ukrainian, and Vietnamese Moonshine models under a permissive open-source license.

Related papers

• AfroXLMR-Comet: Multilingual Knowledge Distillation with Attention Matching for Low-Resource languages [0.19381162067627603]
  We present a novel hybrid distillation approach that combines traditional knowledge distillation with a simplified attention matching mechanism.<n>We evaluate our approach on five African languages: Kinyarwanda, Swahili, Hausa, Igbo, and Yoruba.
  arXiv  Detail & Related papers  (2025-02-25T09:28:47Z)
• Aya Expanse: Combining Research Breakthroughs for a New Multilingual Frontier [72.5652085347547]
  We introduce the Aya Expanse model family, a new generation of 8B and 32B parameter multilingual language models.<n>By leveraging several years of research at Cohere For AI and Cohere, Aya Expanse sets a new state-of-the-art in multilingual performance.<n>Our evaluations on the Arena-Hard-Auto dataset, translated into 23 languages, demonstrate that Aya Expanse 8B and 32B outperform leading open-weight models.
  arXiv  Detail & Related papers  (2024-12-05T15:41:06Z)
• Whisper Finetuning on Nepali Language [0.0]
  This research focuses on making an exhaustive and generalized dataset followed by fine-tuning OpenAI's Whisper models to improve transcription accuracy for the Nepali language. We leverage publicly available ASR datasets and self-recorded custom datasets with a diverse range of accents, dialects, and speaking styles further enriched through augmentation. Our approach outperforms Whisper's baseline models trained on Fleur's dataset, achieving WER reductions of up to 36.2% on the small and 23.8% on medium models.
  arXiv  Detail & Related papers  (2024-11-19T15:55:56Z)
• ML-SUPERB 2.0: Benchmarking Multilingual Speech Models Across Modeling Constraints, Languages, and Datasets [106.7760874400261]
  This paper presents ML-SUPERB2.0, which is a new benchmark for evaluating pre-trained SSL and supervised speech models. We find performance improvements over the setup of ML-SUPERB, but performance depends on the downstream model design. Also, we find large performance differences between languages and datasets, suggesting the need for more targeted approaches.
  arXiv  Detail & Related papers  (2024-06-12T21:01:26Z)
• To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation [16.655022975392992]
  Current multilingual ASR models are compute-intensive and lack proper comprehensive evaluations. We distill knowledge from large teacher models into smaller student variants that are more efficient. Our best-distilled model's overall performance ($45.0$% WER) surpasses that of a SoTA model twice its size.
  arXiv  Detail & Related papers  (2024-06-06T21:11:53Z)
• Efficient Compression of Multitask Multilingual Speech Models [0.0]
  DistilWhisper is able to bridge the performance gap in ASR for these languages while retaining the advantages of multitask and multilingual capabilities. Our approach involves two key strategies: lightweight modular ASR fine-tuning of whisper-small using language-specific experts, and knowledge distillation from whisper-large-v2.
  arXiv  Detail & Related papers  (2024-05-02T03:11:59Z)
• DistilXLSR: A Light Weight Cross-Lingual Speech Representation Model [16.31307448314024]
  We propose DistilXLSR, a distilled cross-lingual speech representation model. By randomly shuffling the phonemes of existing speech, we reduce the linguistic information and distill cross-lingual models using only English data. Our method is proven to be generalizable to various languages/teacher models and has the potential to improve the cross-lingual performance of the English pre-trained models.
  arXiv  Detail & Related papers  (2023-06-02T07:03:06Z)
• Beyond English-Centric Bitexts for Better Multilingual Language Representation Learning [99.42850643947439]
  We show that going beyond English-centric bitexts, coupled with a novel sampling strategy, substantially boosts performance across model sizes. Our XY-LENT XL variant outperforms XLM-RXXL and exhibits competitive performance with mT5 XXL while being 5x and 6x smaller respectively.
  arXiv  Detail & Related papers  (2022-10-26T17:16:52Z)
• How Phonotactics Affect Multilingual and Zero-shot ASR Performance [74.70048598292583]
  A Transformer encoder-decoder model has been shown to leverage multilingual data well in IPA transcriptions of languages presented during training. We replace the encoder-decoder with a hybrid ASR system consisting of a separate AM and LM. We show that the gain from modeling crosslingual phonotactics is limited, and imposing a too strong model can hurt the zero-shot transfer.
  arXiv  Detail & Related papers  (2020-10-22T23:07:24Z)
• Unsupervised Cross-lingual Representation Learning for Speech Recognition [63.85924123692923]
  XLSR learns cross-lingual speech representations by pretraining a single model from the raw waveform of speech in multiple languages. We build on wav2vec 2.0 which is trained by solving a contrastive task over masked latent speech representations. Experiments show that cross-lingual pretraining significantly outperforms monolingual pretraining.
  arXiv  Detail & Related papers  (2020-06-24T18:25:05Z)
• Structure-Level Knowledge Distillation For Multilingual Sequence Labeling [73.40368222437912]
  We propose to reduce the gap between monolingual models and the unified multilingual model by distilling the structural knowledge of several monolingual models to the unified multilingual model (student) Our experiments on 4 multilingual tasks with 25 datasets show that our approaches outperform several strong baselines and have stronger zero-shot generalizability than both the baseline model and teacher models.
  arXiv  Detail & Related papers  (2020-04-08T07:14:01Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.