YouTube Has a Hidden API — Here's How to Use It (No Key, No Quotas)

#api #python #youtube #webdev

Most developers use the YouTube Data API v3. It requires an API key, has strict quotas (10,000 units/day), and charges for heavy usage.

But YouTube.com itself doesn't use that API. It uses something called Innertube — an internal API that powers the YouTube website and mobile apps.

And you can use it too. No API key. No quotas. No OAuth.

What is Innertube?

Innertube is YouTube's internal API. When you visit youtube.com, every video load, every comment section, every search result — all of it goes through Innertube.

The endpoints are public (they have to be — your browser calls them). They just aren't documented.

Get Video Comments (No API Key)

import requests
import json

def get_comments(video_id, max_comments=20):
    """Get comments from a YouTube video using Innertube."""
    url = "https://www.youtube.com/youtubei/v1/next"

    payload = {
        "context": {
            "client": {
                "clientName": "WEB",
                "clientVersion": "2.20240101.00.00"
            }
        },
        "videoId": video_id
    }

    resp = requests.post(url, json=payload)
    data = resp.json()

    # Navigate the response to find comments
    # (simplified — real implementation needs continuation tokens)
    contents = data.get("contents", {})
    print(f"Got response for video {video_id}")
    return data

# Usage
data = get_comments("dQw4w9WgXcQ")

Search YouTube (No API Key)

def search_youtube(query, max_results=10):
    """Search YouTube using Innertube."""
    url = "https://www.youtube.com/youtubei/v1/search"

    payload = {
        "context": {
            "client": {
                "clientName": "WEB",
                "clientVersion": "2.20240101.00.00"
            }
        },
        "query": query
    }

    resp = requests.post(url, json=payload)
    data = resp.json()

    # Extract video results
    results = []
    contents = (data
        .get("contents", {})
        .get("twoColumnSearchResultsRenderer", {})
        .get("primaryContents", {})
        .get("sectionListRenderer", {})
        .get("contents", []))

    for section in contents:
        items = (section
            .get("itemSectionRenderer", {})
            .get("contents", []))
        for item in items:
            video = item.get("videoRenderer", {})
            if video:
                results.append({
                    "title": video.get("title", {}).get("runs", [{}])[0].get("text", ""),
                    "videoId": video.get("videoId", ""),
                    "views": video.get("viewCountText", {}).get("simpleText", ""),
                    "channel": video.get("ownerText", {}).get("runs", [{}])[0].get("text", "")
                })
    return results[:max_results]

# Usage
videos = search_youtube("python tutorial 2024")
for v in videos:
    print(f"{v['title']} — {v['views']}")

Get Video Transcripts (No API Key)

def get_transcript(video_id):
    """Get video transcript/subtitles using Innertube."""
    # First, get the player response
    url = "https://www.youtube.com/youtubei/v1/player"

    payload = {
        "context": {
            "client": {
                "clientName": "WEB",
                "clientVersion": "2.20240101.00.00"
            }
        },
        "videoId": video_id
    }

    resp = requests.post(url, json=payload)
    data = resp.json()

    # Get captions
    captions = (data
        .get("captions", {})
        .get("playerCaptionsTracklistRenderer", {})
        .get("captionTracks", []))

    if captions:
        # Get the first caption track (usually auto-generated)
        caption_url = captions[0]["baseUrl"]
        transcript = requests.get(caption_url).text
        return transcript
    return None

transcript = get_transcript("dQw4w9WgXcQ")
print(transcript[:500] if transcript else "No captions available")

Innertube vs YouTube Data API v3

Feature	Innertube	YouTube Data API v3
API key	Not needed	Required
Quotas	None	10,000 units/day
Rate limits	Soft (be polite)	Strict
Comments	All + replies	Limited
Search	Full	Full
Transcripts	Yes	No (separate API)
Documentation	None (reverse-engineered)	Official docs
Stability	Can change anytime	Versioned

Important Caveats

Innertube is undocumented — YouTube can change it anytime without notice
Be respectful — don't hammer the API. Add delays between requests.
No guarantees — the response format may change
Terms of Service — check YouTube's ToS for your use case

Pre-Built Tools

If you don't want to parse Innertube responses yourself, I built Apify actors that do it:

DEV Community