> ## Content Index
> Fetch the complete content index at: https://community.mozilladatacollective.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# 15 Datasets for Fine-Tuning Whisper on a New Language in 2026
- URL: https://community.mozilladatacollective.com/15-datasets-for-fine-tuning-whisper-on-a-new-language-in-2026/
- Published: 2026-06-09T10:40:51.000Z
- Updated: 2026-06-09T10:40:51.000Z
- Author: Mozilla Data Collective
- Tags: News

## Why fine-tuning Whisper is a dataset problem

[OpenAI's Whisper](https://openai.com/index/whisper/?ref=community.mozilladatacollective.com) changed what's possible in speech recognition: a single multilingual model with strong zero-shot performance across dozens of languages. But "dozens" is the catch. But for the long tail of the world's languages, even for some with tens of millions of speakers, Whisper's out-of-the-box output is unusable in terms of word error rate, the most popular metric.

Fine-tuning fixes this, and the fine-tuning recipe is well understood by now: take a Whisper checkpoint, feed it paired audio and transcripts in your target language, and the model adapts fast. How fast can be striking: the documentation for the [Khmer ASR Cultural Dataset](https://mozilladatacollective.com/datasets/cmpdy0icy00l9nu07zo5hl1m3?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) below reports that adding even a few hundred Khmer speech-text pairs to the training mix measurably lowered Whisper Large V2's character error rate, bringing it down to roughly 8%. 

What’s clear is that the bottleneck isn't the training code but the data. You need audio that's transcribed accurately, licensed in a way that lets you use it, and ideally varied enough in speaker, dialect, and recording condition that your fine-tuned model generalises beyond the training set.

This article lists 15 automatic speech recognition datasets on Mozilla Data Collective that are well suited to Whisper fine-tuning, spanning a range of language families (Dravidian, Bantu, Uralic) and regions (Mesoamerican, South Asian, European). They range from large studio-production corpora to small, phonetically precise field recordings. Several are time-aligned, which is exactly the format Whisper fine-tuning wants. Pick the one in your target language, or combine several related ones to push a whole language family across the usability threshold.

## The 15 datasets

### South Asia

- [**Malayalam Time-Aligned Speech Corpus**](https://mozilladatacollective.com/datasets/cmmno795h009hml07dh7uefvp?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: Community | Licence: CC-BY-NC-4.0 | Size: 1.50 GB, 6 hours | Task: ASR | Format: WAV, SRT
- [**Tamil Time-Aligned Speech Dataset**](https://mozilladatacollective.com/datasets/cmnmyptri02glo107p5cx5por?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: MirasAI | Licence: CC-BY-NC-SA-4.0 | Size: 37.11 MB, 5 hours | Task: ASR | Format: OGG, SRT
- [**Kannada Time-Aligned Speech Corpus**](https://mozilladatacollective.com/datasets/cmnggocgg007ymh079k30st39?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: MirasAI | Licence: CC-BY-NC-SA-4.0 | Size: 355.77 MB, 5 hours | Task: ASR | Format: OGG, SRT

### Southeast Asia

- [**Khmer ASR Cultural Dataset (Version 3 - Part 5)**](https://mozilladatacollective.com/datasets/cmpdy19nn00llnu07y94pzeu4?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: DDD-Cambodia | Licence: CC-BY-SA-4.0 | Size: 33.14 GB, 87 hours | Task: ASR | Format: WAV
- [**Khmer ASR Cultural Dataset (Version 3 - Part 7)**](https://mozilladatacollective.com/datasets/cmpdy0icy00l9nu07zo5hl1m3?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: DDD-Cambodia | Licence: CC-BY-SA-4.0 | Size: 29.93 GB, 81 hours| Task: ASR | Format: WAV
- [**Mandar Spontaneous Speech**](https://mozilladatacollective.com/datasets/cml5e30pd00eskr072e6a4rrh?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: Community | Licence: CC-BY-NC-4.0 | Size: 534.45 MB, 10 hours | Task: ASR | Format: MP3, TSV
- [**Jember Javanese Spontaneous Speech Corpus**](https://mozilladatacollective.com/datasets/cmlgm5a94008kny07nz2intus?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: Universitas Gadjah Mada | Licence: CC-BY-NC-SA-4.0 | Size: 271.65 MB, 10 hours | Task: ASR | Format: MP3, TSV

### African Region

- [**DataTrust Africa: Speech Corpus of Public Radio Recordings from Northern Uganda**](https://mozilladatacollective.com/datasets/cmkfm6xtw00k2nv07oakesnix?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: Community | Licence: NOODL-1.0 | Size: 179.82 MB | Task: ASR | Format: WAV, TSV
- [I**siZulu Second Language Learner Speech Corpus**](https://mozilladatacollective.com/datasets/cmo1y8vrv006ol207lc86hc13?ref=community.mozilladatacollective.com) Steward: Community | Licence: CC-BY-SA-4.0 | Size: 5.26 GB | Task: ASR | Format: WAV, SQLite

### Siberia and Northern Eurasia

- [**INEL Nganasan Speech Corpus**](https://mozilladatacollective.com/datasets/cmn4kxhp8001tnz07jyfvf2qt?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: University of Hamburg | Licence: CC-BY-NC-SA-4.0 | Size: 1.41 GB, 38.5 hours | Task: ASR | Format: TSV, MP3
- [**INEL Kalmyk Speech Corpus**](https://mozilladatacollective.com/datasets/cmn4kxlaj001xnz07a5yugnew?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: University of Hamburg | Licence: CC-BY-NC-SA-4.0 | Size: 138.31 MB, 3 hours | Task: ASR | Format: TSV, MP3
- [**INEL Evenki Speech Corpus**](https://mozilladatacollective.com/datasets/cmn4kxexu001pnz07kn6wr985?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: University of Hamburg | Licence: CC-BY-NC-SA-4.0 | Size: 103MB, 2.6 hours | Task: ASR | Format: TSV, MP3
- [**INEL Dolgan Speech Corpus**](https://mozilladatacollective.com/datasets/cmn4kqzzt0013nu07caxllg3t?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: University of Hamburg | Licence: CC-BY-NC-SA-4.0 | Size: 583.34 MB, 13 hours | Task: ASR | Format: TSV, MP3

### Latin America

- [**Speech Corpus of English Learners from Mexico**](https://mozilladatacollective.com/datasets/cmo29suhm00gjo2078rhmqn3p?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: Community | Licence: CC-BY-SA-4.0 | Size: 2.45 GB, 8 hours | Task: ASR | Format: MP3, TSV
- [**Archivo GELED: Muestra general de audios del cuicateco**](https://mozilladatacollective.com/datasets/cmo29qkrl00g5o207hash6gk2?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model) Steward: Community | Licence: CC-BY-NC-SA-4.0 | Size: 1.85 GB, 3 hours | Task: ASR | Format: WAV, TSV

## From scraped to shared: fine-tuning Whisper on community datasets

Fine-tuning Whisper on a new language is one of the highest-leverage things you can do in speech AI right now: a few hours of well-transcribed audio can take a language from unusable to production-viable. For a step by step guide on how to fine-tune Whisper models using Mozilla Data Collective datasets you can [check out our step-by-step tutorial](https://community.mozilladatacollective.com/fine-tune-a-speech-to-text-model-for-any-language-including-yours/). The 15 datasets above give you audio across a deliberately wide spread of language families: Dravidian (Tamil, Kannada, Malayalam), Bantu (isiZulu), Uralic and Mongolic (Nganasan, Kalmyk, Evenki, Dolgan); and regions: Mesoamerican (Cuicatec), and Southeast Asian (Khmer, Javanese).

Mozilla Data Collective is rebuilding the data ecosystem by keeping communities front and centre and giving the speakers of a language, and the organisations that record and preserve it, a say in how their speech datasets get licensed, credited, and applied. For under-served languages, this is a meaningful shift: away from models trained on whatever material can be scraped and extracted from the internet, and toward systems built on datasets that communities have deliberately shared and contributed on terms they set.

[Browse all Mozilla Data Collective datasets →](https://mozilladatacollective.com/datasets?utm%5Fsource=community.mozilladatacollective.com&utm%5Fmedium=blog&utm%5Fcampaign=fine-tune-whisper-model)

[Get in touch →](mailto:support@mozilladatacollective.com)

[Join Mozilla Data Collective → ](https://mozilladatacollective.com/auth/signup?ref=community.mozilladatacollective.com)