凤希AI - 音频列表预览

If you've ever queued a long multi-segment video, walked away for an hour, and come back to discover segment seven's audio is 40 seconds of silence, you know why FxAiAudioListPreview (凤希AI - 音频列表预览) is worth a slot. It takes a LIST of AUDIO segments and renders a player for each one in the frontend, so you can actually listen before you burn a whole render cycle. Preview audio is one of those things vanilla ComfyUI is strangely thin on; this node is the pack's answer, and it's the audio twin of the image previewers most people already use.

How it works

Simple and robust. It walks the LIST of AUDIO objects, converts each waveform to a mono-appropriate wav, writes it to ComfyUI's temp directory with soundfile, and pushes the file paths to the UI through the standard ui.audio mechanism. The frontend (this pack ships its own small JS for it) then shows a player per segment. What comes back through the 音频数量 output is the count of segments that actually previewed - which doubles as a free "did the list survive the trip" check.

Under the hood it's careful in the ways that matter: it skips anything that isn't a proper audio dict, forces stereo where needed so every player behaves the same, and clips anything beyond two channels. If a segment fails to save it logs and moves on rather than killing the whole preview.

Input

One required input:

• 音频 - a LIST of AUDIO. Feed it the 分段列表-style outputs from this pack's segmenters, or any other audio LIST you've assembled (say with easy-makeAudioList or ComfyUI's list helpers).

Output

• 音频数量 (audio count) - an INT, the number of segments successfully previewed. Handy as a wiring sanity check; most of the time you'll ignore it and just look at the players.

Install

One node in fxai-toolkit (凤希全能节点包), the Chinese MIT-licensed utility suite from 凤希AI built around segmented long-video generation. ComfyUI Manager (search "fxai") or:

cd ComfyUI/custom_nodes
git clone https://github.com/fxai666/fxai-toolkit
# restart ComfyUI

The pack auto-installs soundfile (which this node uses to write wavs) and psutil on first boot. No model downloads.

Gotchas

• It's a preview node, which means it's an output node - it must be on the graph's execution path to run, and it marks the end of a branch. It re-runs every time (like all previews) so don't expect it to be skipped by caching.
• The players are per-segment, which is exactly what you want for checking sync. If you wanted one continuous audition of the whole track, use FxAiAudioMerge instead and preview that.
• A segment that's pure silence still previews - silence is valid audio. If a segment shows up as a flatline, that's the upstream loader or slicer's fault, not this node's.
• Everything's labeled in Chinese: 音频 = audio, 数量 = count.