ai-agent

FireRedASR2S: Multi‑Language Speech & Singing Recognition for Global Startups

Name: FireRedASR2S: Multi‑Language Speech & Singing Recognition for Global Startups
Author: FireRedTeam

Deploy an industrial‑grade ASR pipeline that handles Mandarin (20+ dialects), English, code‑switching, and singing, plus VAD, LID, and punctuation—all in Python.

From the repo · “A SOTA Industrial-Grade All-in-One ASR system with ASR, VAD, LID, and Punc modules. FireRedASR2 supports Chinese (Mandarin, 20+ dialects/accents), English, code-switching, and both speech and singing ASR. FireRedVAD supports speech/singing/music in 100+ langs. FireRedLID supports 100+ langs and 20+ zh dialects. FireRedPunc supports zh and en.”

advanced⏱ 1-2 days💵 Free + LLM API costs

722 stars+7★ this week47 forksPythonGuide quality 8/10 · reviewed Oct 11Updated 6/2/2026100% free · open source

What it is

Use FireRedASR2 to recognize and understand speech from audio.

Real use case example

"A founder wants to analyze customer satisfaction from call center recordings. She uploads the audio to FireRedASR2 to generate transcripts. Then, she trains a machine learning model to find keywords expressing dissatisfaction. Within weeks, she identifies frequent pain points and adjusts her product features to match customer needs."

Best insideSelf-hosted

When to use it

•When you need to transcribe speech or singing in various languages and dialects with high accuracy
•When your application requires robust speech recognition for both speech and music
•When you need to support a wide range of languages, including those with complex dialects and accents

Ready-to-paste prompt

python -m firered.asr.infer --config config.yaml --audio file.wav --output output.txt

Heads up: Make sure you have the required libraries and dependencies installed, including PyTorch and TensorFlow, and that your system meets the minimum requirements specified in the README.md file

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🛠️ Technical setup

Expect 20–40 minutes in a terminal — or let your AI agent drive it.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/fireredasr2s && curl -fsSL https://workflowstacks.com/api/skills/fireredasr2s/claude-skill -o ~/.claude/skills/fireredasr2s/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How FireRedASR2S: Multi‑Language Speech & Singing Recognition for Global Startups works

Codeflow

Free to inspect

FireRedASR2S: Multi‑Language Speech & Singing Recognition for Global Startups is a medium Python project (~8.2k lines across 76 code files). Expect some technical setup — comfortable with a terminal, or ask a developer. Last commit 5 months ago, Apache-2.0 license, no tests found.

Size

Medium codebase

~8.2k lines · 76 code files · ~1 h to skim

Setup

Some technical setup

Comfortable with a terminal? 20–40 min. Otherwise ask a dev.

Runs on

Python

No API keys detected

Python 96%Shell 4%

Where to start reading

1
README.md
Start here — what it does and how to install it
2
pyproject.toml
Dependencies and the commands it exposes
3
fireredasr2s/__init__.py
Inside fireredasr2s/ — the main logic begins here

What's in each folder

fireredasr2s/Core code — the actual logic59 files

examples_infer/Folder27 files

runtime/Folder17 files

assets/Images & static assets5 files

READMENo tests foundDocumentedNo CIApache-2.0 licenseUpdated 5 mo ago

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🛠️ Technical setup

Expect 20–40 minutes in a terminal — or let your AI agent drive it.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/fireredasr2s && curl -fsSL https://workflowstacks.com/api/skills/fireredasr2s/claude-skill -o ~/.claude/skills/fireredasr2s/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

Quick Actions

Details

Creator

FireRedTeam

Language

Python

Do this with a working workflow

Pages we wrote and tested that put a skill like this to work — importable n8n templates, outcome guides and Claude Desktop configs.

Outcome guide

Automate your meeting summaries

A private form takes any transcript (from Zoom, Meet, Otter, Fireflies) and emails you back a structured summary: a TL;DR, the decisions made, and action items with owners.

Outcome guide

Turn meeting notes into a summary email automatically

Paste the transcript into a private form and the workflow writes and emails a concise recap — decisions, owners, open questions — that people will actually read.