EvoLink Sora 2 Pro (Official)

Sora 2 Pro is what Sora 2 becomes when you need the delivery version rather than the draft. Same OpenAI model family, same EvoLink wrapper - but with a quality switch that gets you to 1080p, and a price tag that reflects it. If you're making something that'll actually be watched, this is the node you'll reach for.

What Pro buys you

The regular Sora 2 node tops out at 720p and 12 seconds. Pro adds a quality choice - 720p or 1080p - and the tooltip is refreshingly honest about the economics: 1080p runs about 1.67× the price of 720p. With billing by the second on top, a 12-second 1080p clip is where a video session starts to add up. That's not a bug; it's the real price of these renders surfacing. Budget accordingly: 720p for iteration, 1080p for the final cut.

The mechanism is otherwise identical to Sora 2 - submit to /videos/generations, poll, download before the 24-hour links die, hand back a native VIDEO tensor. Text-to-video and image-to-video both work through the same image input. Same strict moderation, same blunt rule: no real-person photos as reference, and no bypassing it, because there are no local weights to patch.

The inputs that matter

• prompt - required, ~5000 tokens, Chinese-friendly. Shot description, director's-note style.
• aspect_ratio - 16:9 or 9:16. No square, same as Sora 2.
• duration - 4 / 8 / 12 seconds, billed by the second.
• quality - the Pro-only switch: 720p (default) or 1080p. This is the field that separates this node from plain Sora 2.

Outputs

video → Save Video, and that's the whole job. task_id is your charge receipt for the job log at evolink.ai, status_info is the human-readable summary (model, elapsed, credits used), result_urls the raw 24-hour links, response_json the full platform payload for automation. Because the whole pack shares this output shape, you can wire the clip into the Topaz upscale node or a Kling/Seedance edit node downstream without any adapter.

Install and gotchas

Ships in the comfyui-evolink pack. Manager → search EvoLink → Install → fully restart ComfyUI (close the console, not just refresh). Or:

cd ComfyUI/custom_nodes
git clone https://github.com/deeplearning-goethe/comfyui-evolink

Only requests as a dependency; no models, no GPU. API key from evolink.ai/dashboard/keys (sk-), fill once, blank afterward so the key stays out of shared workflow JSON.

The usual suspects apply: node missing after install means you didn't fully restart (or the folder's double-nested); a native VIDEO output means you need a current ComfyUI build. timeout_seconds defaults to 1200 - video is slow, let it cook. Failed and moderation-blocked tasks don't charge, and unchanged reruns hit the ComfyUI cache instead of the API, which on Pro pricing is a difference you'll feel in the job log.