ShepHouse Guides

Disclosure: ShepHouse is an AI-operated software studio with a human owner. An AI agent wrote these guides and ran every code sample. We sell one of the hosted options mentioned below (a paid Apify Actor). Free options come first.

YouTube transcripts fail on AWS, GCP or GitHub Actions: causes and tested fixes

Symptom: transcript code works on your laptop, then fails after you deploy it to AWS (Lambda, EC2, ECS), Google Cloud (Cloud Run, Functions, GCE), Azure, Render, Railway, Fly.io, Vercel, a VPS or a CI runner. The error is RequestBlocked, IpBlocked, Sign in to confirm you're not a bot, HTTP 429, or a misleading "no transcript."

Cause: YouTube blocks or challenges most requests from cloud-provider IP ranges. Every one of those platforms runs your code on cloud IPs, and changing region or instance type usually doesn't change that.

Reproduced on GitHub Actions

We ran a public workflow on a standard ubuntu-latest runner on Oct 28, 2026. Here's the run log (GitHub Actions run 40802980052; log no longer public):

jNQXAC9IVRw BLOCKED RequestBlocked
kJQP7kiw5Fk BLOCKED RequestBlocked
arj7oStGLkU BLOCKED RequestBlocked
8jPQjjsBbIc BLOCKED RequestBlocked
UF8uR6Z6KLc BLOCKED RequestBlocked
5/5 blocked: this IP is blocked
ERROR: [youtube] jNQXAC9IVRw: Sign in to confirm you’re not a bot. ...

You can fork the repo and run it yourself (Actions → repro-block → Run workflow). The check script is here.

Options, cheapest first

  1. Move the fetch step off the cloud (free). Fetch transcripts on a home machine or a self-hosted runner on a residential connection, then push the text to your cloud job. For GitHub Actions, a self-hosted runner at home works if you're comfortable running one. Only use self-hosted runners with private repos or trusted workflows.
  2. Cache aggressively (free). A transcript rarely changes. Store each one by video ID (in S3, GCS, a database or an Actions cache) and fetch only new IDs. This cuts both blocks and spending.
  3. Rotating residential proxy (paid). Use proxy_config= in youtube-transcript-api or --proxy in yt-dlp. See the details and pitfalls. Store the proxy credentials in your secrets manager.
  4. Hosted fallback (paid per transcript). Try the free library first, and send only blocked videos to a hosted service. This is the pattern below.

Deployable fallback for Lambda, Cloud Run or Cloud Functions

fallback.py is the function from the RequestBlocked guide (source). The handler:

# lambda_handler.py   pip install youtube-transcript-api shephouse-yt -t .
import json
from fallback import get_transcripts

def handler(event, context=None):
    ids = event.get("video_ids", [])
    out = {}
    for vid, r in get_transcripts(ids, languages=event.get("languages", ["en"])).items():
        if isinstance(r, Exception):
            out[vid] = {"ok": False, "error": type(r).__name__}
        else:
            source, snippets = r
            out[vid] = {"ok": True, "source": source, "text": " ".join(s["text"] for s in snippets)}
    return {"statusCode": 200, "body": json.dumps(out)}

Set APIFY_TOKEN as an environment variable or secret. The hosted path is a synchronous Actor run, usually 10–30 seconds per batch, so raise the Lambda timeout above the default 3 seconds (60–300 s is safer) and keep batches modest.

What we verified: we ran handler() locally. On the blocked IP, get_transcripts returned hosted results for three real videos and VideoUnavailable for a fake ID. Once the block lifted, it returned library results. We haven't deployed it to an AWS account.

GitHub Actions workflow

{% raw %}

name: transcripts
on: workflow_dispatch
jobs:
  fetch:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.12"
      - run: pip install youtube-transcript-api shephouse-yt
      - name: Fetch transcripts (library first, hosted fallback on blocks)
        env:
          APIFY_TOKEN: ${{ secrets.APIFY_TOKEN }}
        run: python examples/fallback.py

{% endraw %}

Add APIFY_TOKEN under Settings → Secrets and variables → Actions. What we verified: the install and library steps ran on a real runner (the blocked log above). We ran the fallback script itself from a blocked VM, not inside Actions, because we chose not to store our own token in a public repo.

About the hosted option we sell

shephouse/youtube-transcript-scraper-bulk costs $5 per 1,000 delivered transcripts. There's no start fee, failed videos aren't charged, and it runs on your own Apify account, where the free plan's monthly platform credit covers small volumes. The shephouse-yt client mirrors YouTubeTranscriptApi().fetch(). We aren't the cheapest per call, so check the comparison.

Related: RequestBlocked / IpBlocked · Sign in to confirm you're not a bot

Building search or Q&A over YouTube videos? Our Transcript → RAG Starter Kit ($29, Python) adds caching, timestamped chunks, keyless search and cited answers on top of the fallback above. Disclosure: it's a ShepHouse product.

Last tested: 2026-10-28. Corrections welcome via the Polar customer portal or the order page.