ByteDance Text to Video

ByteDance Text to Video is the most barebones entry point into ByteDance's Seedance line inside ComfyUI: a text prompt, a resolution, an aspect ratio, a duration - and out comes a video. No image to feed, no frames to match, no asset registration to worry about. It's the 1.x-generation API node (default model seedance-1-0-pro-fast), the text-only sibling of ByteDance Image to Video, and the one you use when you just want to see what Seedance does with an idea before you commit to a more elaborate first-frame or reference workflow.

Same partner-node rules as everything else in this family: it ships with ComfyUI core, it's a cloud call through Comfy's API (no local model, no VRAM), auth is your logged-in Comfy account, and each generation is billed through your credits. If you want the current 2.x generation line with dialogue steering and a Fast/Mini cost ladder, that's ByteDance2TextToVideoNode - this node is the simpler, older, still-working option.

The inputs

• prompt - the text prompt used to generate the video.
• model - defaults to seedance-1-0-pro-fast, with other Seedance 1.x models available.
• resolution and aspect_ratio - output size and shape.
• duration - 3 to 12 seconds.
• seed (optional) - used for generation; 0 = random.
• camera_fixed (optional) - asks the platform to append a "fix the camera" instruction to your prompt. The tooltip is honest about the ceiling: it does not guarantee the effect.
• watermark (optional) - add an "AI generated" watermark.
• generate_audio (optional) - ignored for every model except seedance-1-5-pro. Tick it on the default and nothing changes.

What comes out

One VIDEO output.

Gotchas

• Prompt-only means prompt responsibility. With no image anchor, the whole composition lives in your text - be explicit about subject, environment, lighting, and motion or the model fills in the gaps its own way.
• generate_audio needs the right model. Same trap as the i2v node: it silently does nothing unless you're on seedance-1-5-pro.
• camera_fixed is a nudge. Plan for motion; don't trust the toggle to hold the shot.
• Credits spend like anywhere in this family. Draft short and low-res, then scale up for the take you keep.

If you're assembling a ByteDance workflow, this node is the fast sketchpad - and when a text-only draft looks right, that's the signal to move up to the first-frame or 2.x tiers for the real render.