⏰Ding Dong Text

pc ding dong text (⏰Ding Dong Text) is the chatty sibling in the PC-ding-dong pack. Where the plain Ding Dong node plays a chime, this one speaks your words out loud the moment execution reaches it. "Batch 7 of 20 done." "Your upscale finished." "Something broke, probably." It turns a long silent render into something with a running commentary, which is exactly the small quality-of-life win that keeps people from staring at the progress bar.

How it works

Same architecture as its sibling, one crucial difference: the voice isn't an audio file - it's synthesized by your browser. When the node runs, the backend sends a websocket message (pc.play_ding_dong_text) with your text, pitch, rate and volume, and the pack's frontend JavaScript hands it to the browser's built-in Web Speech API (speechSynthesis.speak). There's no TTS model, no server-side synthesis, no model download - the voice is whatever your operating system and browser have installed.

That's the whole mechanism, and it's both the charm and the limitation. Charm: zero setup, zero cost, and your text never leaves your machine for some TTS vendor. Limitation: voice quality is entirely your browser's business. If your OS has a decent installed voice, this sounds fine; on a barebones setup you might get something that belongs in a GPS from 2009.

One nice detail: the frontend tries to auto-detect the language of your text using a small JS library called franc (loaded from a CDN) and sets the utterance's language so you get a matching voice if one exists. If the CDN is unreachable, it quietly falls back to the default voice and language - no failure, no error.

Inputs

• text - STRING, default "Hello, World!". Whatever it should say.
• pitch - 0.1–2, default 1. Robot-voice territory above 1.5.
• rate - 0.1–10, default 1. Speech speed, not a sampler setting; 1 is normal talking pace, 2 is auctioneer.
• volume - 0–1, default 1. Note the scale: this node's volume is normalized 0–1, unlike Ding Dong's 0–100.
• anything - required any-type passthrough, same as its sibling. Wire in your finished image or latent so it speaks at exactly the right moment; it comes back out output unchanged.

Output: output, the passthrough. It's an output node, so it can sit at the end of a workflow as the "say the thing, mark done" terminal.

Installing it

Same pack, same one-liner. ComfyUI Manager → ComfyUI-PC-ding-dong, or:

cd ComfyUI/custom_nodes
git clone https://github.com/lgldlk/ComfyUI-PC-ding-dong

Restart and you're done. No dependencies, no model files, no API key.

The gotchas

All the browser-side caveats from Ding Dong apply, plus one extra: if your browser has no usable speech voices - rare on desktop, but a real thing on minimal Linux builds - nothing gets spoken at all. Check your OS voice settings first, because the node will happily "work" and say nothing. And the pack itself is unmaintained (a couple of years stale as of 2026); the community's go-to replacement for the whole "tell me when the workflow is done" idea is comfyui-ping, which keeps the browser-tab approach and adds more settings. This one's a fun, zero-dependency way to get the same thing today.