SAVRN
Search Contact SAVRN

Open-weight model · Speech recognition

wav2vec2-large-voxrex-swedish

by National Library of Sweden / KBLab KBLab/wav2vec2-large-voxrex-swedish

Finetuned version of KBs VoxRex large model using Swedish radio broadcasts, NST and Common Voice data. Evalutation without a language model gives the following: WER for NST + Common Voice test set (2% of total sentences) is 2.5%.

Parameters315M
Context—
Weights2.5 GB
Licensecc0-1.0
AccessOpen weights
Monthly Downloads1.9M

Runs On

What it takes to serve wav2vec2-large-voxrex-swedish (315M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.6 GB 0.8 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.3 GB 0.4 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.2 GB 0.2 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on wav2vec2-large-voxrex-swedish

The hardware question answers itself here. KBLab, at the National Library of Sweden, fine-tuned its VoxRex large model on Swedish radio, the NST database and Common Voice, and it needs 0.8 GB of memory at 16-bit and 0.2 GB at 4-bit. The cheapest setup we price, one 192 GB MI300X at $1.85 an hour on demand, only pays if Swedish transcription is one of many jobs on the card; size by concurrent transcriptions, not by fit.

The license reads CC0 1.0, and our page carries no summary, so read the text from the publisher before you build on it. Two things to check. Input must be sampled at 16 kHz. And training ran in two stages, 120,000 updates on NST plus Common Voice, then 20,000 on Common Voice alone, which the publisher says hurt performance on the combined set; the paper is arXiv:2205.03026.

SAVRN Research, 2026-09-18

Model Card

By National Library of Sweden / KBLab, published under cc0-1.0, revision ca70e31c06a2.

Wav2vec 2.0 large VoxRex Swedish (C)

Finetuned version of KBs VoxRex large model using Swedish radio broadcasts, NST and Common Voice data. Evalutation without a language model gives the following: WER for NST + Common Voice test set (2% of total sentences) is 2.5%. WER for Common Voice test set is 8.49% directly and 7.37% with a 4-gram language model.

When using this model, make sure that your speech input is sampled at 16kHz.

Update 2022-01-10: Updated to VoxRex-C version.

Update 2022-05-16: Paper is is here.

Performance*

*Chart shows performance without the additional 20k steps of Common Voice fine-tuning

Training

This model has been fine-tuned for 120000 updates on NST + CommonVoice and then for an additional 20000 updates on CommonVoice only. The additional fine-tuning on CommonVoice hurts performance on the NST+CommonVoice test set somewhat and, unsurprisingly, improves it on the CommonVoice test set. It seems to perform generally better though [citation needed].

Usage

Read the full model card (292 words)

Configuration

Architecture
Wav2Vec2ForCTC
Layers
24
Hidden size
1,024
Feed-forward size
4,096
Attention heads
16
Vocabulary size
46
Model type
wav2vec2

Identity and Version

Repository
KBLab/wav2vec2-large-voxrex-swedish
Publisher
National Library of Sweden / KBLab
Task
Speech recognition
Modality
Audio
Library
transformers
Parameters
315M parameters
Languages
sv
Revision
ca70e31c06a2617bf7fe3b4fb5d387d2b19b2983
First published
2022-03-02
Last updated
2025-05-22

Files and Weights

11 files, 2.5 GB in total. The weights are 2 files totalling 2.5 GB in bin, safetensors.

Weights2 files · 2.5 GB
Configuration3 files · 2.1 KB
Tokenizer2 files · 632 B
Documentation1 file · 3.4 KB
Other2 files · 339.7 KB
Repository1 file · 791 B
Every file
FileTypeSizeSHA-256
model.safetensorsWeights1.3 GB d09f24c48fdb
pytorch_model.binWeights1.3 GB 7138b3f9c570
config.jsonConfiguration1.8 KB —
preprocessor_config.jsonConfiguration212 B —
special_tokens_map.jsonConfiguration85 B —
README.mdDocumentation3.4 KB —
chart_1.svgOther189.6 KB —
comparison.pngOther150.1 KB —
.gitattributesRepository791 B —
tokenizer_config.jsonTokenizer211 B —
vocab.jsonTokenizer421 B —

License and Download

License
cc0-1.0
Access
Open weights, no gate
Download size
2.5 GB
Download from National Library of Sweden / KBLab

Released by National Library of Sweden / KBLab through its official repository on Hugging Face. Read the license.

Built From

  • Described by arXiv:2205.03026
  • Trained on (disclosed) NST_Swedish_ASR_Database
  • Trained on (disclosed) P4
  • Trained on (disclosed) common_voice

Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

BenchmarkConditionsResultReported byRevisionDate
Common Voice Task Speech RecognitionMetric Test WERComparison conditions not established 8.49 KBLab
Publisher reported		 Evaluated revision not stated —

Memory Requirements

PrecisionWeights in memory
As published2.5 GB
16-bit0.6 GB
8-bit0.3 GB
4-bit0.2 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About wav2vec2-large-voxrex-swedish

How much GPU memory does wav2vec2-large-voxrex-swedish need?

About 0.8 GB at 16-bit and 0.2 GB at 4-bit: the weights (315M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run wav2vec2-large-voxrex-swedish on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

What license is wav2vec2-large-voxrex-swedish released under?

cc0-1.0, as its publisher declares it. Read the license text before commercial use.

Similar Models

Model · Speech recognition

wav2vec2-xls-r-300m-cs-250

Vojtěch Drábek

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the commonvoice 8.0 dataset as well as other datasets listed below. It achieves the following results on the evaluation set: The eval.py script results using a LM are: Fine-tuned facebook/wav2vec2-large-xlsr-53 on Czech using the Common Voice dataset. When using this model, make sure that your speech input is sampled at 16kHz. The model can be used directly (without a language model) as follows: The model can be evaluated using the attached eval.py script: The Common Voice 8.0 train and validation datasets were used for training, as well as the following datasets: - Šmídl, Luboš and Pražák, Aleš, 2013, OVM – Otázky…

Open weights apache-2.0 315M parameters transformers
View model

This model for Croatian ASR is based on the facebook/wav2vec2-xls-r-300m model and was fine-tuned with 300 hours of recordings and transcripts from the ASR Croatian parliament dataset ParlaSpeech-HR v1.0. If you use this model, please cite the following paper: Nikola Ljubešić, Danijel Koržinek, Peter Rupnik, Ivo-Pavao Jazbec. ParlaSpeech-HR -- a freely available ASR dataset for Croatian bootstrapped from the ParlaMint corpus. http://www.lrec-conf.org/proceedings/lrec2022/workshops/ParlaCLARINIII/pdf/2022.parlaclariniii-1.16.pdf Evaluation is performed on the dev and test portions of the ParlaSpeech-HR v1.0 dataset. There are multiple models available, and in terms of CER and WER, the…

Open weights 315M parameters transformers
View model

Model · Speech recognition

w2v-xls-r-uk

Smoliakov

ATTENTION! Metrics (float16) using evaluate library with batchsize=1

Open weights apache-2.0 315M parameters transformers
View model

Model · Speech recognition

koelsch-wav2vec2-ipa

Chem

IPA phoneme recognition for Kölsch (Ripuarian German, Cologne). Fine-tuned from facebook/wav2vec2-xls-r-300m with a CTC head over a 48-symbol IPA inventory. This is also the forced-alignment model. Its frame-level CTC posteriors drive torchaudio.functional.forcedalign to produce Praat TextGrids with word and phone tiers — see notebook 9 in the repository. Trained during the CIF Tandem Fellowship at IfL-Phonetik, University of Cologne, on Alles Kölsch (Bhatt & Lindlar 1998) — 4,670 utterances, 4.5 hours of spontaneous Cologne dialect speech from 105 speakers aged 10–88 across 49 neighbourhoods. Kölsch is Ripuarian German. It has no public speech dataset and no standardised spelling: 94.1 %…

Open weights cc-by-nc-sa-4.0 315M parameters transformers
View model

Model · Speech recognition

romanian-wav2vec2

Théo Gigant

You can test this model online with the Space for Romanian Speech Recognition The model ranked TOP-1 on Romanian Speech Recognition during HuggingFace's Robust Speech Challenge: This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the Common Voice 8.0 - Romanian subset dataset, with extra training data from Romanian Speech Synthesis dataset. Without the 5-gram Language Model optimization, it achieves the following results on the evaluation set (Common Voice 8.0, Romanian subset, test split): The architecture is based on facebook/wav2vec2-xls-r-300m with a speech recognition CTC head and an added 5-gram language model (using pyctcdecode and kenlm) trained on the Romanian…

Open weights apache-2.0 315M parameters transformers
View model

Model · Speech recognition

wav2vec2-large-xls-r-300m-Urdu

Abid Ali Awan

A fine-tuned XLS-R 300M CTC model for Urdu automatic speech recognition. It transcribes 16 kHz mono audio and includes an optional 5-gram KenLM decoder. Best reported result: 39.89% WER / 16.70% CER with KenLM decoding on the Urdu Common Voice 8.0 test set. See the Kaggle evaluation notebook for a reproducible example. The repository contains a 5-gram KenLM language model. The Kaggle notebook evaluates a five-sample streaming smoke test from fixie-ai/commonvoice170 (ur, test). Results are reported on the Urdu test split of Mozilla Common Voice 8.0. The language-model row is the model-card score; compare each result only with the same decoding strategy. To reproduce language-model evaluation…

Open weights apache-2.0 315M parameters transformers
View model