From 5dfbfcbdc04e9a502a0c396ba9ac8febcdb2b557 Mon Sep 17 00:00:00 2001
From: pere <per@capia.no>
Date: Tue, 19 Dec 2023 08:48:21 +0100
Subject: [PATCH] =?UTF-8?q?Adding=20Norwegian=20Bokm=C3=A5l=20and=20Norweg?=
 =?UTF-8?q?ian=20Nynorsk?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Adding Wav2Vec2-models for Norwegian Bokmål and Norwegian Nynorsk. The models are testet together with WhisperX, and works great. For Bokmål I have added the 1B model, even if I see fairly little difference between that and the 300M model. For Norwegian Nynorsk only a 300M exist.The quality of the Wav2Vec models are also reported here: https://arxiv.org/abs/2307.01672
---
 whisperx/alignment.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/whisperx/alignment.py b/whisperx/alignment.py
index cd7f8ec..8294983 100644
--- a/whisperx/alignment.py
+++ b/whisperx/alignment.py
@@ -53,6 +53,8 @@ DEFAULT_ALIGN_MODELS_HF = {
     "hi": "theainerd/Wav2Vec2-large-xlsr-hindi",
     "ca": "softcatala/wav2vec2-large-xlsr-catala",
     "ml": "gvs/wav2vec2-large-xlsr-malayalam",
+    "no": "NbAiLab/nb-wav2vec2-1b-bokmaal",
+    "nn": "NbAiLab/nb-wav2vec2-300m-nynorsk",
 }