Seedance 2.0 Mini just dropped in ComfyUI, finally the cost of cranking out drafts in bulk is way down.

The most annoying thing about making short video ads is endlessly iterating through a pile of garbage footage—every single one using the full-fat model just burns through cash like crazy. Now in ComfyUI, ByteDance’s Seedance 2.0 video nodes (text-to-video, first/last frame, reference image to video) all have a “Mini” option you can pick directly, plus they come with dedicated templates.

Mini’s positioning is crystal clear: same nodes, same prompt handling, lower cost—the trade-off is resolution capped at 720p (native 1080p and 4K still need the full 2.0 and 2.0 Fast). For someone like me who needs to batch-test shots and scripts, it’s perfect—same visual feel, but you can churn out a bunch of drafts on the cheap first, then pick the winners and render them in HD for the final cut.

It reads prompts as one continuous visual narrative, generating rhythmic multi-shot sequences where characters stay consistent throughout—great for product stories with characters. To get started, just update ComfyUI to the latest version, load a Mini template from the template panel, or add a Seedance 2.0 node and switch the model to Mini, set resolution to 480p or 720p, throw in your prompt, and let it rip.

Yeah, the draft-first approach makes sense — only upscale the final picks. Who’s gonna waste full power on junk images anyway?

Sitting here waiting for someone to actually test it out.

Multi-shot character consistency is a real need. Before, with other models, the face would change by the second shot.

Honestly, 720p is more than enough for feed ads. It’s not like anyone’s gonna zoom in on their phone anyway.