zhenzhen-video-g-omni-flash（旧版兼容）

What this node actually is

This is the previous version of zhenzhen's omni-flash video node, and the display name says it out loud: 旧版兼容 - "legacy compatibility." When the author rebuilt omni-flash as v2 with an explicit mode dropdown, every workflow that had already saved the old node would have broken on reload. So the old one stayed, frozen, so your existing graphs keep working. If you're starting fresh, use Comfly_zhenzhen_video_g_omni_flash_lowprice_v2 and don't look back. This page exists for the moment you load an old workflow and wonder what this thing is.

What it can still do

Being "old" doesn't mean useless. The v1 node is a one-node-fits-most omni-flash box, and it covers four jobs:

• Text to video - prompt, ratio, done.
• Image to video / multi-image reference - up to 16 image inputs.
• Video editing - feed a clip into input_video, or paste a URL into video_url.
• Task continuation - put a previous run's extend_from_task_id in and continue from where it left off.

The trade-off against v2: no mode selector, no resolution choice (locked to 720p), and no duration control. The old node figures out what you want from which inputs you bothered to connect. That's convenient, and it's also exactly why v2 exists - implicit behavior is hard to debug.

The inputs

Only two are required: prompt and ratio (default 16:9, with adaptive, 4:3, 1:1, 3:4, 9:16 and 21:9 on offer). Everything else is optional wiring. The ones worth knowing:

• api_config - the shared Seedance 2.0 Low Price Settings node, same as every other node in this category.
• input_video / video_url - two ways to give it a source clip, and they're mutually exclusive. Connect one or the other; the code throws if you do both.
• extend_from_task_id - continuation, and it's also exclusive with video_url. Pick one job per run.
• skip_error - when on, failures return an empty error result instead of killing the graph.

Outputs and install

Outputs are the standard zhenzhen quartet: video (ComfyUI VIDEO), video_url, task_id, and response (JSON). Install is the same for the whole pack - ComfyUI Manager, search "Comfyui-zhenzhen", or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

then restart and let the requirements install.

Gotchas

The mutual-exclusion checks are the thing most likely to bite: input_video and video_url both connected → error, and it tells you so. Also remember this is a cloud call, not a local model - the "video editing" job uploads your source clip to zhenzhen's server, so it's subject to whatever the upstream provider filters. If a run 500s, rerun it; the README's standing advice for these API nodes is that the provider is usually the one having a bad day.