FireRedLID

FireRedLID

Identifies which language is spoken in an audio clip, across 100+ languages and 20+ Chinese dialects and accents. The language ID module of FireRedASR2S, for speech developers.

Made by FireRedTeam on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: FireRedLID is the spoken language identification module of FireRedASR2S, an all-in-one speech recognition system. You give it audio, and it says which language is being spoken. It covers 100+ languages and 20+ Chinese dialects and accents.

Who it suits: It suits speech developers who need to know the language of a recording before they transcribe or route it.

Title: Who runs it
Industry | Who uses it | How they use it | Example
Software | Speech developers | Detect the language before choosing a recognizer | Send Mandarin clips to a Mandarin recognizer
Media | Teams sorting audio archives | Label recordings by language | Group a folder of clips by language
Research | Speech researchers | Compare language identification accuracy | Score the model on FLEURS data

If you only work in one known language, you can skip it.

Where to use it:

• Python on Windows, macOS or Linux, with the FireRedASR2S code installed
• A GPU, since the card's sample turns on GPU use, though the setting can be changed
• Larger FireRedASR2S pipelines, where this is the language step

Title: Jobs it does
Use case | Who | What they get back
Labelling a clip's language | Developers | A language, a confidence score and the clip length
Batch sorting | Archive teams | One result per audio file in a list

How to set it up:

• Python on Windows, macOS or Linux
• The FireRedASR2S code from the project's GitHub repository, with its dependencies
• WAV files to test with

1. Clone the FireRedASR2S repository on GitHub and install its dependencies.
2. Download the model files.
3. Run the Python sample.

pip install -U huggingface_hub
hf download FireRedTeam/FireRedLID

How to use it: Build a config, load the model by name, and pass a list of clip ids with a list of WAV paths. The call returns one result per clip.

from fireredasr2s.fireredlid import FireRedLid, FireRedLidConfig

batch_uttid = ["hello_zh", "hello_en"]
batch_wav_path = ["assets/hello_zh.wav", "assets/hello_en.wav"]

config = FireRedLidConfig(use_gpu=True, use_half=False)
model = FireRedLid.from_pretrained("FireRedTeam/FireRedLID", config)

results = model.process(batch_uttid, batch_wav_path)
print(results)

Title: Calls in the sample
Tool | What it does
FireRedLidConfig | Sets options such as GPU use and half precision
FireRedLid.from_pretrained | Loads the model by its Hugging Face name
model.process | Identifies the language of each listed clip

Example: You pass two short clips, one named hello_zh and one named hello_en. The card's sample output labels the first as Mandarin and the second as English, each with a confidence of 0.996, plus the clip length and a speed figure.

For newcomers and experts: If you are new, language identification means a program listens to a recording and tells you what language it is. The first small thing to try is the two-clip sample from the card.

If you know the field, the card measures utterance-level accuracy. FireRedLID scores 97.18 on the FLEURS test across 82 languages, against 92.91 for SpeechBrain and 79.41 for Whisper. The trade-off is coverage: the figures cover 82 and 74 languages, not all 100+ it supports.

Clone the code and install it
Load FireRedLID by name
Pass clip ids and WAV paths
Read the language and confidence

Good to know:

Title: Accuracy on the FLEURS test across 82 languages
FireRedLID: 97.18%
SpeechBrain: 92.91%
Whisper: 79.41%

• Other sets: 92.07 on the CommonVoice test and 88.47 on Chinese dialect and accent sets.
• Setup: the model alone is not enough, since the card asks you to clone its code first.
• License: apache-2.0.
• Cost to run: no API key, account or paid service is named.
• Data it reaches: it reads the audio files you pass in and downloads the model from Hugging Face.

On Hugging Face, 13 November 2026

• Task: audio classification
• Downloads in the last 30 days: 622
• Likes: 29
• License: apache-2.0
• Languages: en, zh
• Quantized versions: 1
• Fine-tunes: 1
• Tags: voxlingua107, audio, automatic-speech-recognition, audio-classification

Links

• Model: https://huggingface.co/FireRedTeam/FireRedLID
• Files: https://huggingface.co/FireRedTeam/FireRedLID/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download FireRedTeam/FireRedLID

Then follow the usage in its model card. Nothing in this kit runs by itself.