Concurrent Submit | zhenzhen-video-g-omni-flash（旧版兼容）

Check the display name and you'll see the tell: "旧版兼容" - "old-version compatible." This is the legacy edition of the pack's Google omni-flash video node, kept alive so saved workflows don't break. Its Concurrent Submit wrapper runs it through the 10-slot video pool, but the actual model behavior here is the old omni-flash: 720p only, one resolution option, no mode dropdown. If you're building a new workflow, you probably want the v2 node instead. If you're resurrecting an old JSON, this is why it still loads.

omni-flash is Google's fast video model - the README tracks it arriving as omni_flash-10s at "1 credit per 10 seconds," i.e. cheap and fast, aimed at quick drafts. The legacy node keeps the older, simpler interface for the older API contract.

How the legacy node differs from v2

The v2 node added a mode dropdown (text / frame / reference_images / reference_video), resolution up to 4k, and a 4/6/8/10-second selector. The legacy version has none of that structure: it's a flat prompt + 720p + ratio + a grab bag of optional inputs. What it does keep is the interesting stuff for old workflows:

• extend_from_task_id - a string task id that continues a previous generation. The "task continuation" feature lives here in the legacy node.
• input_video / video_url - feed a video for editing/reference.
• image1–image16 - up to sixteen reference images.
• ratio - adaptive, 16:9, 4:3, 1:1, 3:4, 9:16, 21:9.

So it's a t2v/i2v/edit/continue all-in-one with no mode switch - the API infers what you're doing from what you connected. That's the old way, and it's genuinely convenient for a saved workflow that was built before the v2 split.

Inputs and outputs

• prompt - the scene. Multiline.
• resolution - 720p. The only option. This is the compatibility ceiling.
• ratio - seven shapes, adaptive default.
• api_config - zhenzhen-seedance2-low-price-api-settings (domestic key).
• seed - cache-control only.

Output: task (COMFLY_VIDEO_FUTURE) → Concurrent Collect Videos (10) → VIDEO tensors + status.

Install

ComfyUI Manager → search "Comfyui-zhenzhen" → install → restart, or:

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

No model downloads, API key required.

The honest advice

• New work: use v2. The mode dropdown is clearer, 4k exists, and the new API contract is where the author is actually maintaining things. Don't build fresh on the legacy node just because it's in the menu.
• Old work: don't touch anything. If a workflow you saved back when omni-flash was 720p-only still loads and runs, leave it. This node exists precisely so you don't have to migrate.
• extend_from_task_id is the one feature to remember. Continuing from a prior task id is how you chain longer videos out of a fixed-length model, and this is where that lived.

One thing to note about the whole omni-flash family: Google's "fast" tier is a draft model, and 720p legacy output is a draft tier of a draft tier. Use it to test scenes and motion ideas, not to ship deliverables.