Turn prompts, stills, and mixed references into 2–30 second clips at 480p–1080p, with optional audio and thinking mode.
Wan 3.0 renders one continuous clip from 2 to 30 seconds — long enough for a setup and a payoff, short enough to test a move at 2 seconds. Thinking spends extra interpretation on stacked camera, blocking, and timing; it does not change length, resolution, or credit cost.
Whole-second clips. You set the length; the UI does not auto-pad or shorten.
480p, 720p, or 1080p. Image-to-video follows the first frame’s aspect ratio.
Reads stacked camera, blocking, and timing in the prompt before it renders the take.
Images, video, and audio. Video and audio each total 15 seconds or less.
Shot 01
Use this when the scene exists in a prompt, not a still — a camera move that has to land, a beat that needs a middle, not a five-second loop you will stitch later.
Thinking does not pick the duration for you and does not add credits. You still set the length. It only spends more interpretation on the prompt you wrote.
Shot 02
Use this when you already know the start and end composition — a product that should finish on a hero angle, a character who walks from a doorway to a window.
One image is a start frame. A second image is the last frame the model should land on, not a style board. If you have a character sheet and no destination frame, reference-to-video is the better mode.
Shot 03
Use this when the face, the performance, or the voice should come from files you already have, not from adjectives in the prompt.
SparkVid will reject an empty reference set — you need at least one image, clip, or audio file. Put identity in images, performance in video, voice or bed in audio, and say how they relate in the prompt. PDFs and webpages are not accepted here.
All three run from the same prompt box and credit balance. Choose per shot, not per account.
Reach for this when
You want the cheapest 1080p 30-second take on SparkVid, a 2-second minimum for tests, and a thinking toggle for stacked camera and blocking. Audio is optional.
2–30s · 480p / 720p / 1080p · 2 / 4 / 8 credits per second · 10 images, 5 videos, 5 audio tracks (15s media cap each).
Reach for this when
The shot is built from a pile of references, not a short prompt. On SparkVid that budget is 30 images, 10 videos, and 10 audio tracks, with 30 seconds of reference media — double Wan’s media cap — at a higher per-second rate.
4–30s · 480p / 720p / 1080p · 6 / 12 / 24 credits per second when no reference video is attached.
Reach for this when
Delivery needs native 2K (2560×1440) and stereo audio in one pass — product type, packaging, or a voiced spot that has to survive a close inspect. Clips stop at 15 seconds.
5–15s · 480p / 768p / 2K · native stereo · up to 9 images, 3 clips, and 3 audio tracks.
How this generator behaves — length, cost, references, and when to switch models.
Set a length, a resolution, and whether thinking is worth the wait. Credits are shown before you generate.