Streamlit-WebRTC

Yuichiro Tachibana (@whitphx)

Build a complete real-time audio and video web app. Using only Python.

Add live camera and microphone input to a Streamlit app, process the media in Python, and return transformed video or audio through WebRTC. Display text, metadata, and metrics with Streamlit elements. Everything lives in one Python script, with no frontend code. Streamlit-WebRTC supplies the browser component and handles WebRTC transport.

❤️

Quick demo

Build real-time computer vision with one callback in ~20 lines of Python

app.pyPython

import av
import streamlit as st
from streamlit_webrtc import webrtc_streamer
from ultralytics import YOLO

model_name = st.selectbox(
    "YOLO model",
    ["yolo11n.pt", "yolo11s.pt", "yolov8n.pt"],
)

with st.spinner(f"Loading {model_name}..."):
    model = YOLO(model_name)

def process(frame: av.VideoFrame) -> av.VideoFrame:
    image = frame.to_ndarray(format="bgr24")
    result = model(image, verbose=False)[0]
    annotated = result.plot()
    return av.VideoFrame.from_ndarray(
        annotated, format="bgr24"
    )

webrtc_streamer(
    key="detection",
    video_frame_callback=process,
    media_stream_constraints={"video": True, "audio": False},
)

Streamlit application showing YOLO model selection and object detection boxes over a live webcam feed

Get started

Start with a Streamlit app

What is Streamlit?: A Python web UI framework for building interactive apps without frontend code.
What is Streamlit-WebRTC?: A custom component that plugs into an existing Streamlit app.
Install: uv add streamlit streamlit-webrtc
Run: uv run streamlit run app.py

How it works

How Python reaches the browser in real time

Turn Python models into user-facing apps without adding a second language or implementing WebRTC. Streamlit-WebRTC provides three media roles: sinks consume input for Streamlit outputs, filters transform and return media, and sources generate media. Compose them independently for audio and video.