RenderZero Video Prompt Bridge

The video sibling of the Prompt Bridge, and where every RenderZero video generation actually starts. You pick a resolution tier, a frame count, and a frame rate in the RenderZero UI; this node converts that into concrete numbers your video model can chew on - prompt, width, height, frame count, frame rate, and seed. It exists because video models are unforgiving about dimensions in a way still-image models aren't: get the canvas wrong and your 81-frame clip comes out tiled, stretched, or outright refused.

How it works

Same architecture as the image bridge - a fixed lookup table plus a pass-through - but the table is built for video models specifically. The three resolution_tier values map to 16-divisible sizes that keep Wan and friends happy: 480p 16:9 resolves to the Wan-safe 832×480, and 1080p 16:9 to 1920×1088 - note the 1088, not 1080. That extra 8 on one axis is the difference between a model that runs and one that errors on a divisibility check, and it's exactly the kind of detail you'd otherwise discover the hard way. The node also outputs the frame rate twice - frame_rate_int and frame_rate_float - because some video models want an integer and some want a float, and this way one bridge feeds both without a conversion node in between.

The inputs that matter

• positive_prompt (STRING, multiline) - passed through to the model untouched.
• aspect_ratio - 1:1, 3:4, 4:3, 9:16, 16:9, 21:9.
• resolution_tier - 480p, 720p, 1080p. Default 720p, which is the sensible place to start for most local setups.
• frame_count (INT, default 81) - total frames, not seconds. 81 is a common Wan starting point.
• frame_rate (FLOAT, default 24) - clamped to a minimum of 0.1 if you undershoot.
• seed (INT) - what RenderZero randomizes between generations unless locked.

The outputs - positive_prompt, width, height, frame_count, frame_rate_int, frame_rate_float, seed - wire straight into your video model's prompt, size, frame-count, frame-rate, and seed inputs, with the model deciding which frame-rate variant it wants.

Installing it

The whole pack installs as one repo. ComfyUI Manager, search "renderzero-comfyui-nodes," Install, restart:

cd ComfyUI/custom_nodes
git clone https://github.com/jamesstothard/renderzero-comfyui-nodes renderzero_bridge_nodes

Restart and it's live. No requirements.txt, no model downloads - pure Python on ComfyUI's own helpers, which means none of the dependency pain this ecosystem is famous for.

Common issues

The workflow rules matter here more than anywhere: video workflows must contain exactly one Video Prompt Bridge, one Video Start Frame Bridge, and one Video Output - and must not contain any of the image-side bridges. Mix them and RenderZero rejects the graph. Also remember the bridge only produces numbers; if your video model ignores them and generates at some other resolution, check that you actually wired width and height into the model's size inputs rather than the sampler's. And a practical one: 1080p at 81 frames is a lot of VRAM. The node will happily hand you 1920×1088; whether your GPU can chew it is a separate conversation.