P-Video

P-Video is Pruna's text-to-video model in the Runware pack, and its standout setting is right there in the node: settings.audio defaults to on. Prompt a scene and you get motion and a synthesized soundtrack, returned as a native VIDEO. That's a bigger deal than it sounds - the KB's closed-source-models essay calls native audio the one capability open video models still lack, and this node hands it to you per-run without any post-production.

The rest of the node is a well-equipped text-to-video: frame images for image-to-video, size presets, duration, fps, and a few genuinely useful quality settings most video nodes don't expose.

What you actually set

• positivePrompt - required. Describe the scene and the motion; with audio on, a line about the soundscape ("waves, distant seagulls") helps.
• frameImages - IMAGE socket for image-to-video. P-Video does solid i2v.
• settings.audio - default True: include audio in the output. This is the headline feature; turn it off if you only need silent footage.
• settings.promptUpsampling - default True: automatically enhance the prompt before generating. It's a black box that expands your prose into better conditioning; keep it on unless you want exact prompt control.
• settings.draft - default False: lower-resolution preview generation. The classic iterate-cheaply-then-commit move, Pruna-flavored.
• size - 720p and 1080p presets, default 720p (16:9).
• duration - default 5 seconds, plain integer.
• fps - 24 or 48. 48 is smooth and slow; 24 is the standard.
• audio - a STRING taking a UUID or URL if you want to supply your own track instead of the generated one.
• seed, numberResults - the usuals.

Output: one video socket, type VIDEO.

Install

One-time pack install: ComfyUI Manager → search Runware → install → restart, or

cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt

API key into ComfyUI Settings → Runware API key or RUNWARE_API_KEY. Light deps, no weights on disk.

Where people get burned

• Audio-on is a cost lever. Generating a soundtrack isn't free; if you're iterating, flip settings.audio off for drafts and on for the final.
• Prompt upsampling changes your words. If you've carefully tuned a prompt and the output drifts, promptUpsampling is the suspect. It's usually a win; it's not always invisible.
• safety.mode defaults to none. Opt into content checking if you want it.
• Video bills stack. Duration, fps 48, and multiple results each push the per-run cost; the title bar shows the damage.

The verdict: P-Video is the most complete of the Pruna video nodes - audio on by default, i2v support, draft mode for cheap iteration. If you're comparing it to the PixVerse line in the same pack, think of it as the "does everything, sounds included" option.