Seedance 1.0 Pro Fast

Seedance 1.0 Pro Fast is the same ByteDance text-to-video model as Seedance 1.0 Pro, with the pipeline tuned for speed instead of peak fidelity. It exists because video inference is expensive and slow, and nobody wants to burn full-price credits discovering their prompt doesn't work. Think of it as the sketch layer: get the composition, motion, and pacing right in Fast, then render the keeper in the full model. In the pack it's a sibling node under Runware/Video/bytedance, and the wiring is identical.

How it works

Same request shape as Seedance 1.0 Pro - prompt in, optionally frameImages for image-to-video, native VIDEO out - but the backend runs an accelerated variant of the generation pipeline. The trade is the standard speed/fidelity one: quicker turnaround and cheaper-feeling iteration, with output that's close but not identical to the full model's best. The practical workflow is two-stage: iterate here, then rebuild the winning prompt in Runware_bytedance_seedance_1_0_pro for the final render.

The inputs that matter

The schema is nearly identical to 1.0 Pro, with one meaningful difference:

• size - a wider menu: 22 presets that add 720p to the 480p/1080p ladder. So Fast isn't just faster, it also gives you a mid-res rung the base node skips.
• positivePrompt (required) - same advice as the base node: write for motion and scene.
• frameImages - start frames for image-to-video.
• duration (1.2–12s), seed, numberResults (1–4) - same controls, same defaults.
• providerSettings.bytedance.cameraFixed - the camera lock that makes shots feel deliberate rather than drifting. On for locked compositions.
• safety toggle + safety.mode (none/fast/full), outputFormat (MP4/WEBM/MOV), outputQuality (20–99) - the usual suspects.

Wiring it up

Output is video, straight into a VIDEO preview/save node. Because the node is a drop-in twin, you can keep two copies of the same prompt - one wired into Fast for iteration, one into the full Pro node for finals - and swap the wire when a take gets approved.

Install and setup

Same one-time pack install: ComfyUI Manager → "Runware" → install → restart, or clone plus pip install -r ComfyUI-Runware/requirements.txt. Light deps (runware-sdk, pillow, soundfile), API key via Settings / RUNWARE_API_KEY / runware auth login, and funded credits - the ~$20 minimum top-up complaint applies to every node here, and it stings most on video models.

Troubleshooting

• "Fast" still takes a while - it's faster than the full model, not instant. Long durations and 1080p eat the gain; iterate at 480p and short durations.
• Final render differs from the approved Fast take - expected. The Fast preview is a guide for structure and pacing, not a pixel-for-pixel promise. Nail the prompt in Fast, accept that the final is its own render.
• Video returns as a file path - the old-ComfyUI degradation again; check the output folder.
• Which node do I actually add? - for a real project, both. Fast for iteration, the full Pro node for the keeper.

This is the honest answer to "can I afford to iterate on video?" - yes, if you iterate here. It's the same model family, half the frustration, and the natural entry point to ByteDance's video line.