Skip to content
ModelListedExternal
Download

Whisper Tiny 24lang

A plain openai/whisper-tiny (unmodified architecture) fine-tuned on the 24 languages of the CC0 Whispered corpus. This is the matched scientific control for burakaydinofficial/whisper-tiny-mla-24lang — trained...

Sourcehuggingface.co/burakaydinofficial/whisper-tiny-24lang/tree/ff4c9dc9a7ded07600efcc9175bf35da7f0b78ae
KindText modelBase modelopenai/whisper-tinyVersionmodelCgrade

Model source

Kind
Text model
File size
72.0 MB
Base model
openai/whisper-tiny
Version
vmodel
Updated
Sep 27, 2026
Downloads
9
Source
Hugging Face

Source description

Whisper-Tiny-24Lang — 24 languages fine-tune, standard architecture (scientific control)

A plain openai/whisper-tiny (unmodified architecture) fine-tuned on the 24 languages of the CC0 Whispered corpus. This is the matched scientific control for burakaydinofficial/whisper-tiny-mla-24lang — trained identically, minus the MHA→MLA conversion — published so the MLA conversion cost is independently reproducible. No custom code: loads directly in transformers, and — being a plain unmodified Whisper — is convertible for faster-whisper / CTranslate2 / whisper.cpp via their standard converters.

from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor
model = AutoModelForSpeechSeq2Seq.from_pretrained("burakaydinofficial/whisper-tiny-24lang")   # no trust_remote_code
processor = AutoProcessor.from_pretrained("burakaydinofficial/whisper-tiny-24lang")

Reproduce the conversion cost

Evaluate this control and whisper-tiny-mla-24lang on CommonVoice-17 (scripts/validate.py in the code repo); the per-language difference is the conversion cost reported on the MLA card and in docs/results/.

Results (CommonVoice-17 test, greedy, Whisper normalization + Arabic folding; CER for th/zh/ja)

Reconstructed numbers. This control's per-language WER/CER in the table are reconstructed as (the MLA model's absolute score − the measured paired conversion cost), not from a separate evaluation of this checkpoint; an independent re-eval may differ by a few tenths. The paired conversion cost is the directly measured quantity.

Langthis control
en28.1 WER
de42.1 WER
es29.5 WER
fr44.8 WER
it43.8 WER
pt43.1 WER
ru40.6 WER
nl38.8 WER
pl47.3 WER
id53.0 WER
tr52.1 WER
hi45.9 WER
ms49.5 WER
sv-SE57.9 WER
th32.0 CER
zh-CN34.7 CER
cs67.2 WER
vi54.4 WER
fi66.9 WER
el63.7 WER
da73.7 WER
ja48.8 CER
nn-NO93.1 WER
ko71.3 WER

Encoder frozen during fine-tuning; 15,000 steps, warmup+cosine, fp16. Read-speech domain (CommonVoice + FLEURS-validated). "Compression cost" does not apply to this unconverted control.

Sources

1 source
Hugging FacePrimaryVersion ff4c9dc9a7ded07600efcc9175bf35da7f0b78ae
huggingface.co/burakaydinofficial/whisper-tiny-24lang/tree/ff4c9dc9a7ded07600efcc9175bf35da7f0b78ae

Verified Aug 30

Model artifacts

1 artifact

model.safetensors

safetensors · 72.0 MB · SHA-256 6bdacecb0b10…9e58 · Hugging Face

Download file

Source excerpts

2 excerpts
ReadmeAug 30huggingface.co/burakaydinofficial/whisper-tiny-24lang

--- license: apache-2.0 base_model: openai/whisper-tiny datasets: - burakaydinofficial/Whispered language: [en, de, es, fr, it, pt, ru, nl, pl, id, tr, hi, ms, sv, th, zh, cs, vi, fi, el, da, ja, "no", ko] pipeline_tag: automatic-speech-recognition library_name: transformers tags: [whisper, multilingual, speech-recognition, fine-tuned, control] --- # Whisper-Tiny-24Lang — 24 languages fine-tune, standard architecture (scientific control) A **plain `openai/whisper-tiny`** (unmodified architecture) fine-tuned on the 24 languages of the CC0 [Whispered](https://huggingface.co/datasets/burakaydinofficial/Whispered) corpus. This is the **matched scientific control** for [`burakaydinofficial/whisper-tiny-mla-24lang`](https://huggingface.co/burakaydinofficial/whisper-tiny-mla-24lang) — trained identically, minus the MHA→MLA conversion — published so the MLA **conversion cost is independently reproducible**. No custom code: loads directly in transformers, and — being a plain unmodified Whisper — is convertible for faster-whisper / CTranslate2 / whisper.cpp via their standard converters. ```python from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor model = AutoModelForSpeechSeq2Seq.from_pretrained("burakaydinofficial/whisper-tiny-24lang") # no trust_remote_code processor = AutoProcessor.from_pretrained("burakaydinofficial/whisper-tiny-24lang") ``` ## Reproduce the conversion cost Evaluate this control **and** `whisper-tiny-mla-24lang` on CommonVoice-17 (`scripts/validate.py` in the [code repo](https://github.com/burakaydinofficial/Whispered)); the per-language difference is the conversion cost reported on the MLA card and in `docs/results/`. ## Results (CommonVoice-17 test, greedy, Whisper normalization + Arabic folding; CER for th/zh/ja) > **Reconstructed numbers.** This control's per-language WER/CER in the table are reconstructed as (the MLA model's absolute score − the measured paired conversion cost), not from a separate evaluation of this checkpoint; an independent re-eval may differ by a few tenths. The paired conversion cost is the directly measured quantity. | Lang | this control | |---|---| | en | 28.1 WER | | de | 42.1 WER | | es | 29.5 WER | | fr | 44.8 WER | | it | 43.8 WER | | pt | 43.1 WER | | ru | 40.6 WER | | nl | 38.8 WER | | pl | 47.3 WER | | id | 53.0 WER | | tr | 52.1 WER | | hi | 45.9 WER |...

NoteAug 30huggingface.co/burakaydinofficial/whisper-tiny-24lang

Source context: 9 downloads · 0 likes · Pipeline automatic-speech-recognition · Library transformers · Repo burakaydinofficial/whisper-tiny-24lang