Reference guide
Choosing a YouTube player_client
When a YouTube download throws HTTP Error 403 or Requested format is not available, the advice you'll find everywhere is to "switch the player client" with --extractor-args "youtube:player_client=...". Sometimes it works. Here's what that flag actually changes, how to read which client yt-dlp is using, which clients return formats right now, and why you shouldn't bake a client name into your config and forget it.
What a "player client" is
YouTube doesn't have one API. The web page, the iOS app, the Android app, the TV interface and the embedded player each ask YouTube's servers for a video through a different internal client, and each client gets back a slightly different set of formats and stream URLs. yt-dlp mimics several of these clients and merges the results. The player_client extractor arg tells it which clients to impersonate.
This matters because YouTube rolls out restrictions (SABR-only streaming, PO-token requirements, 403s) per client, at different times. When your default set comes back broken, a client YouTube hasn't clamped down on yet may still return working URLs — that's the entire reason this flag helps.
See which clients yt-dlp is already using
Before you change anything, look at what it does by default. The verbose log names each client as it queries it:
yt-dlp -v -F "https://www.youtube.com/watch?v=jNQXAC9IVRw" 2>&1 \
| grep -i "player API JSON"
# [youtube] jNQXAC9IVRw: Downloading tv samsung player API JSON
# [youtube] jNQXAC9IVRw: Downloading visionos player API JSONThat is the current default set on a recent build (seen on yt-dlp 2026.10.26, 2026-11-17) — it changes between releases, which is exactly why updating is the first fix, not a flag.
The syntax
Pass a comma-separated list. You can name clients outright, or start from the maintained default set and subtract one with a - prefix:
# force specific clients, in order of preference
yt-dlp --extractor-args "youtube:player_client=tv,web_safari,mweb" URL
# keep yt-dlp's default set but drop one problem client
yt-dlp --extractor-args "youtube:player_client=default,-web_embedded" URL
# list clients with -F first to see what each one returns
yt-dlp --extractor-args "youtube:player_client=mweb" -F URLThe default,-client form is usually the better move: you keep yt-dlp's current, maintained client list and only remove the one that's misbehaving, instead of pinning a single client that may be the next one to break.
Which clients work right now (and why that's a trap)
Here's the honest state of it. Running -F against the same stable public video on yt-dlp 2026.10.26 (2026-11-17, residential connection):
| player_client | Result |
|---|---|
default (no flag) | Full format table |
default,-web_embedded | Full format table |
mweb | Works, but a reduced set (one muxed 240p format) |
web_safari | No video formats found! |
tv | The page needs to be reloaded |
Notice that web_safari and tv— clients forum threads were recommending only weeks ago — return nothing today. This table will be wrong soon, and that's the point: never treat a specific client name as the fix. Run -F with a couple of candidates, use whichever returns the formats you need today, and expect to redo it. If you keep a client flag in a config file, add a note to re-check it.
Do these before you reach for a client flag
Switching clients is a stopgap. The durable fixes address why the default set broke in the first place:
- Update yt-dlp.The default client list and every client's workaround change almost every release. A build from a few weeks ago is the most common reason a client 403s.
yt-dlp -U, oryt-dlp --update-to nightlyif stable is still broken. See update yt-dlp. - Make sure a JS runtime is present. Without one (Deno by default), yt-dlp falls back to a reduced client set — often the SABR-only ones. This bites Docker and servers most: fix "No supported JavaScript runtime" in Docker.
- Supply a PO token if the debug log asks for one. Some surviving format URLs now require a GVS PO token — see forcing SABR streaming & 403 and run a PO token provider on a server.
If no client works on a server
The classic trap: you find a client that works on your laptop, deploy it, and the same command still 403s on an AWS/GCP/Hetzner box — no matter which client you force. A player_client flag can't change where your request comes from, and YouTube flags datacenter IP ranges. When every client fails only on the server, it's the IP, not the client — see yt-dlp works locally but not on my server.
Servers & agents
Works on your laptop but not on the server?
Paste a link to see what a hosted API returns from a clean residential IP — no cookies, proxies, or runtime to maintain on the box.