Sidebar research · compiled 23 April 2026

Proxy options for ~35K
YouTube metadata calls

Anchor job: yt-dlp metadata, 16-way parallel, ~50KB/response, ~1.75 GB total

A pricing and feasibility scan of every IP-rotation route I could think of for a 35K-request, one-off-ish YouTube metadata job. The short version: don't buy proxies for this — use the official YouTube Data API v3, which finishes the work in 7% of one day's free quota. Everything else is here for completeness, and for the day the API isn't enough.

AI-compiled research · prices not personally negotiated
01 · TL;DR

Four picks, four reasons

If you only read the cards: the official API wins on price, simplicity, and CAPTCHA risk. The proxy field has one cheap-enough option for this size of job (Webshare). Most of the rest is built for adversaries YouTube metadata isn't.

Cheapest viable
YouTube Data API v3
$0 · 700 of 10,000 free units
Native batching: 50 IDs per videos.list call. 35K IDs = 700 calls = 7% of one day's free quota. Returns the same fields you currently parse from yt-dlp.
Best balance (if proxy)
Webshare residential
~$2.45 PAYG · $1.40/GB promo
Cheapest verified per-GB residential rate. Works with yt-dlp via a single --proxy gateway URL. Decodo at $3.50/GB (~$6) is a reasonable second.
Lowest setup friction
yt-dlp --proxy gateway
One env var · any rotating provider
Webshare, Decodo, IPRoyal, Bright Data all expose a single rotating-gateway URL. No proxy lists, no rotation code. Set HTTPS_PROXY and re-run.
Probably overkill
Bright Data Web Unlocker, Zyte, ScraperAPI, ScrapingBee, Apify
$0.001–$0.075 per request
Per-request scraping APIs aimed at JS-heavy, anti-bot targets. YouTube metadata via yt-dlp doesn't need any of that machinery.
Honest take up top: you're not blocked. yt-dlp at 4.3 req/s from one residential IP is well under YouTube's pain threshold for innertube traffic. The right move is the YouTube Data API v3 if you want bulletproof + faster — otherwise keep going. Don't pay for proxies for a 35K one-off.
02 · Cost comparison

Every option, priced for this job

All costs scaled to ~35K requests ≈ 1.75 GB of response data. Prices verified 23 April 2026 from vendor pages or 2026-dated third-party reviews. Per-GB and per-request models are not comparable in marketing copy — the per-request column ($/req) absorbs that gap. Green rows are recommended for this job; honey rows are reasonable; muted rows are over-spec.

Provider Type Headline price Cost for 35K reqs Free trial yt-dlp setup Notes
YouTube Data API v3 verified Official API Free, 10K units/day $0 (700 units) n/a not via yt-dlp; requests + key Batches 50 IDs/call. Requires API key.
Webshare residential approx Rotating residential $1.40/GB (promo, ~half off) ~$2.45 10 free proxies forever --proxy gateway Promo flagged on site early 2026.
Decodo (Smartproxy) approx Rotating residential $3.50/GB PAYG, $2/GB at scale ~$6 3-day, 100 MB --proxy gateway 115M+ IPs, good docs.
IPRoyal residential approx Rotating residential $7/GB entry, $1.75/GB bulk ~$12 (entry) none reliably --proxy gateway Non-expiring traffic.
Oxylabs residential approx Rotating residential $4/GB PAYG, $3.49/GB Advanced ~$7 7-day --proxy gateway Aimed at enterprise volume.
SOAX residential approx Rotating residential $4/GB PAYG ~$7 $1.99 / 400 MB / 3 days --proxy gateway Trial is paid but cheap.
Bright Data residential approx Rotating residential $5.88/GB starter, ~$8.40 at 10 GB ~$15 7-day enterprise --proxy gateway Largest pool; overkill here.
NetNut residential approx Rotating residential ~$15/GB entry, $1.59/GB enterprise ~$26 (entry) trial on request --proxy gateway Worst small-volume pricing.
Webshare datacenter approx Rotating DC $2.99/mo for 100 IPs $2.99/mo 10 IPs free --proxy rotation list DC IPs get more YouTube CAPTCHAs.
Bright Data Web Unlocker approx Scraping API $1.50–$12.60/CPM PAYG $50–$440 trial credit swap URL, not yt-dlp YouTube isn't a "hard" target; wasted spend.
ScraperAPI approx Scraping API ~$0.001/credit, JS = 5–25 credits ~$50–$100 5K free credits not yt-dlp; HTTP frontend Premium target multipliers eat budget.
ScrapingBee approx Scraping API 5 credits/req default; $49/mo = 150K ~$11.50 (no JS) 1K free not yt-dlp; HTTP frontend Default JS render = 5× cost.
Zyte API unverified Scraping API $0.06–$16.08 / 1K successes $2–$560 $5 free not yt-dlp; SDK Tier opaque without account.
Apify YouTube actor approx Pay-per-result $0.0005/video (apidojo) ~$17.50 $5/mo free actor run, not yt-dlp Actors range $0.0005–$0.0024/video.
Confidence: verified Google's own docs (C1) approx vendor pricing page or 2026 review (C2) unverified couldn't reach canonical figure without an account (C3)
03 · YouTube Data API v3

The boring answer

This is the right answer for the job and it's worth belabouring, because most people reach for proxies first.

  • Endpoint: GET https://www.googleapis.com/youtube/v3/videos?part=contentDetails,snippet&id=ID1,ID2,…,ID50&key=…
  • Quota cost: 1 unit per call regardless of how many IDs (up to 50). 35K IDs ÷ 50 = 700 calls = 700 units. Daily free quota is 10,000 units. The job finishes in one day, using 7% of free quota.
  • Fields you currently parse from yt-dlp: id, duration, channel — all in contentDetails.duration (ISO 8601) and snippet.channelTitle / snippet.channelId. Direct match.
  • Multi-key rotation: if ever needed, separate Google Cloud projects each get their own 10K/day; trivial round-robin. You don't need it for 35K.
  • Setup cost: 5 minutes — create GCP project, enable YouTube Data API v3, generate API key, restrict by IP / HTTP referer.
  • When this isn't enough: data the API doesn't expose — per-request playback URLs, age-gated content, deleted-but-cached videos, streaming details yt-dlp infers. For "id, duration, channel" it's enough.

The cost-per-request gap between this ($0) and the cheapest paid proxy ($2.45 at Webshare) is small in absolute terms but infinite in ratio. The real benefit of the API is no rate-limit surprises and no CAPTCHA risk, ever.

04 · Residential proxies

If you insist on rotation

All of these work with yt-dlp via yt-dlp --proxy http://user:pass@gateway:port — the provider rotates exit IPs server-side per request, so no proxy list is needed. yt-dlp accepts HTTP/HTTPS/SOCKS5 proxies natively (--proxy); no built-in --proxy-file exists yet.

  • Webshare — cheapest verified ($1.40/GB residential, promo). Free tier of 10 datacenter IPs forever. Pool 80M+ IPs. For 1.75 GB, roughly $2.45 PAYG-style or grab the $3.50/mo 1 GB plan twice. Best value for this job if going paid.
  • Decodo (formerly Smartproxy) — $3.50/GB PAYG, drops to $2/GB at 1 TB. 115M+ IPs. 3-day trial. Reasonable second choice.
  • IPRoyal — $7/GB entry (high), but traffic never expires — useful if this might be a yearly job. 32M+ IPs.
  • Oxylabs / SOAX / NetNut — all in the $4–$15/GB entry range. Aimed at enterprise; small jobs pay tourist prices. Skip for 1.75 GB.
  • Bright Data — biggest, best, most expensive at small volume ($5.88/GB starter rate). KYC and dashboard friction make this objectively wrong for a one-off personal job.
05 · Scraping APIs

What you don't need

These bundle proxy rotation + headless browsing + retries. You don't need any of this for YouTube metadata. They're priced for sites that fight back hard (Amazon, Google SERPs, Cloudflare-protected). Listed for completeness.

  • ScraperAPI — credits scale with target difficulty; YouTube would likely classify as the "premium" multiplier. Effective $0.005–$0.04/req.
  • ScrapingBee — 5 credits/req default (JS on); ~$0.0016/req on the $49 Freelance plan if JS is off, ~$0.008 if not.
  • Bright Data Web Unlocker — $1.50–$12.60/CPM. For 35K reqs: $50–$440. Massively over-engineered for this.
  • Zyte API — tiered $0.06–$16.08/1K. Self-classifies the target; YouTube's tier isn't publicly stated, can't quote a final cost without an account.
  • Apify — $0.0005/video on apidojo's actor = $17.50 for 35K. Fine if you want a hosted runtime, but yt-dlp already works.
06 · Free / DIY

The clever-looking dead ends

All four are worse versions of the YouTube Data API.

  • Tor — free, slow (single-digit req/s shared circuit), and YouTube aggressively rate-limits Tor exits. Don't.
  • Cloudflare Workers free tier — 100K reqs/day. You could write a Worker that proxies your yt-dlp requests through Cloudflare's IP pool. CF egress IPs are well-known and YouTube may treat them as datacenter. Untested for this; experimental.
  • AWS Lambda — Lambda IPs rotate within a region. ~$0.20/million invocations + tiny egress. Effectively $0 for 35K. But again: YouTube treats AWS ranges as datacenter and will CAPTCHA faster than your home IP, not slower.
  • Free VPN rotation — manual, fragile, the IPs are abused; not worth the effort.

The DIY answer that actually works is the YouTube Data API. Everything else here is a worse version of it.

07 · My take for this job

Two real options

You have ~35K IDs, 16 workers, no current blocks, and a one-off-ish workload. Don't buy proxies. Two real options:

  1. Keep going as-is. At 4.3 req/s the run finishes in ~2.2 hours. If you hit a CAPTCHA, pause for an hour and resume. Total spend: $0. The right call if you're already most of the way through.
  2. Switch to YouTube Data API v3. Rewrite the fetcher in ~50 lines of Python: collect IDs, chunk to 50, hit videos.list, parse contentDetails.duration (ISO 8601 → seconds) and snippet.channelTitle. Done in one HTTP-bound run with no rate-limit risk at all (700 of 10,000 units). The right call if this will be re-run periodically or if you want zero block risk.

If you genuinely want a proxy as a fallback safety net for the current yt-dlp run: Webshare residential is the only one priced low enough to make sense at this scale (~$2.45 expected). Set it as HTTPS_PROXY only when retrying failed IDs; don't route all 35K through it.

What I'd flag: cost-per-request math punishes you for tiny responses. A 50KB metadata response on a $5/GB residential plan costs ~$0.00025 — but on a $1/CPM scraping API it's $0.001, 4× more. The two pricing models look comparable in marketing copy and aren't.

08 · Confidence & unknowns

What's solid, what isn't

Where each number sits on the verified→guess axis. Vendor pricing pages change; promo rates (especially Webshare's $1.40/GB) can end without notice — re-check before purchase.

09 · Sources

Where the numbers came from

Pricing pages and one yt-dlp issue. All accessed 23 April 2026.

  1. developers.google.com/youtube/v3/determine_quota_cost
  2. brightdata.com/pricing/proxy-network/residential-proxies
  3. oxylabs.io/pricing/residential-proxy-pool
  4. decodo.com/proxies/residential-proxies/pricing
  5. iproyal.com/pricing/residential-proxies/
  6. webshare.io/pricing
  7. soax.com/pricing
  8. scraperapi.com/pricing
  9. scrapingbee.com/pricing
  10. docs.zyte.com/zyte-api/pricing.html
  11. brightdata.com/pricing/web-unlocker
  12. apify.com/apidojo/youtube-scraper
  13. github.com/yt-dlp/yt-dlp/issues/10718