faster-whisper-51-african-languages

CTranslate2 (float16) conversion of Sunbird/asr-whisper-51-african-languages — Whisper large-v3 fine-tuned for ASR in 51 African languages — for use with faster-whisper.

A different compute_type (e.g. int8_float16, ~1.6 GB VRAM) can be selected at load time and is quantized on the fly.

IMPORTANT: language codes

The fine-tune reuses Whisper's built-in language-token slots for African languages, and faster-whisper only accepts standard Whisper language codes. To transcribe a given language, pass the faster-whisper code from the table below (also machine-readable in language_map.json). Do not rely on automatic language detection.

Language	Sunbird code	pass to faster-whisper as language=
Acholi	ach	su
Afrikaans	afr	af
Akan	aka	jw
Amharic	amh	am
Ateso	teo	bs
Bambara	bam	ba
Bemba	bem	haw
Berber	ber	tt
Chichewa	nya	km
Dagaare	dga	bo
Dagbani	dag	tl
English	eng	en
Ewe	ewe	my
French	fra	fr
Fulani	ful	lb
Hausa	hau	ha
Igbo	ibo	sa
Ikposo	kpo	uz
Kabyle	kab	mt
Kalenjin	kln	ht
Kanuri	kau	nn
Kikuyu	kik	tk
Kinyarwanda	kin	ps
Kwamba	rwm	sq
Lendu	led	lo
Lingala	lin	ln
Luganda	lug	sd
Lugbara	lgg	yi
Luhya	luy	be
Lumasaba	myx	ka
Luo	luo	tg
Lusoga	xog	eu
Malagasy	mlg	mg
Ndebele	nbl	oc
Nigerian Pidgin	pcm	mr
Oromo	orm	pa
Rukiga	cgg	as
Rukonjo	koo	fo
Runyankole	nyn	si
Ruruuli	ruc	gl
Rutooro	ttj	ne
Shona	sna	sn
Somali	som	so
Sotho	sot	kk
Swahili	swa	sw
Thur	lth	gu
Tswana	tsn	mn
Wolof	wol	hy
Xhosa	xho	is
Yoruba	yor	yo
Zulu	zul	br

Usage

import json
from faster_whisper import WhisperModel
from huggingface_hub import hf_hub_download

repo = "Sunbird/faster-whisper-51-african-languages"
lang_map = json.load(open(hf_hub_download(repo, "language_map.json")))

model = WhisperModel(repo, device="cuda", compute_type="float16")

segments, info = model.transcribe(
    "audio.wav",
    language=lang_map["lug"],   # Luganda
    task="transcribe",
    beam_size=5,
    vad_filter=True,
    condition_on_previous_text=False,
)
for segment in segments:
    print(f"[{segment.start:.2f} -> {segment.end:.2f}] {segment.text}")

Notes

• The source repo's reference pipeline normalizes input features (do_normalize=True in WhisperProcessor); faster-whisper's feature extractor does not apply this normalization. Validate WER on your target languages against the Transformers baseline before production use.
• See the base model card for training data, per-language WER/CER, and evaluation methodology.