Core capability
What the Wan 3.0 image-to-video form exposes
One required start frame, one optional end frame, a prompt, and six controls: resolution (480P, 720P, 1080P, 2K, 4K), duration from 3 to 30 seconds, ratio (adaptive, 16:9, 4:3, 1:1, 3:4, 9:16), a sound toggle, prompt enhance and a seed. Images must be JPEG, PNG, BMP or WEBP; the official limits are 240 to 8,000 px per side, no more than 8:1 aspect and 20 MB. Alibaba Cloud’s documentation treats the first frame as strictly the first frame of the video, which is why a clean, well-lit still matters more than any prompt word.
The pass rate and what stops the 9%
Production window: the 21 days to 9 September 2026 unless a 60-day window is named. Rates are rounded down. 182 Wan 3.0 image-to-video jobs from 61 creators: 166 finished (91%), 11 were stopped by the upstream review and 9 of those 11 were rejections of the rendered clip rather than the input. Median render time was 182 seconds. The remaining failures were transient provider errors and were refunded. For comparison, the Wan 2.2 Spicy image-to-video route finished 98% of 16,059 jobs with zero review stops, so the price of Wan 3.0’s audio, length and end-frame control is roughly one blocked clip in sixteen.
Start frame rules that decide the result
Use a still where the face is visible and sharp, the hands are in frame or clearly out of it, and the background has depth. Generate the still on Seedream 5.0 Lite or Qwen Image if you do not have one; both run uncensored here. Crop the still to the ratio you will render at, because adaptive ratio follows the image and a mismatched crop produces letterboxing or a re-composed frame. Avoid stills with motion blur or heavy filters; Wan 3.0 inherits them for the entire clip.
Versions and current access
Wan 3.0 image-to-video is available in Makeartly Studio with start-frame and optional end-frame slots; the CTA opens that route when the live catalogue confirms it.
Evidence
The end frame trick for controlled scenes
The optional end frame tells Wan 3.0 where the motion must arrive. Give it a second still of the same character in the finishing pose (same lighting, same framing) and describe the transition in the prompt: “she lowers herself from standing to kneeling over eight seconds, camera holding steady.” This is the most reliable way to get a specific position change without the anatomy errors that the review then flags. Keep the two frames within the same ratio and within a plausible eight to fifteen seconds of motion.
Deployment filtering and adult-content boundaries
Makeartly does not add a separate mainstream adult-content refusal layer for verified adult creators; this deployment-level statement does not mean the upstream official model is uncensored or guarantee prompt acceptance. Adult workflows are for adults only. CSAM and all under-18 sexual content are prohibited. Consent and likeness authorization are required, and non-consensual intimate content, unauthorized real-person deepfakes, and sexual impersonation are prohibited.
Cost ladder for an image-to-video session
The quote is per second at your tier: 480P 100, 720P 220, 1080P 440, 2K 660, 4K 990 credits per second at review time. A sensible session is three 480P 3-second tests (300 each) to lock the motion, then one 720P or 1080P final at the duration you need. Audio is free. If you only need a silent 5-second loop, the Wan 2.2 Spicy route costs a fraction of that and has no review stops; save Wan 3.0 for clips that need sound, an end frame or more than 15 seconds.
Use cases
You need a specific position change, not a loop
Use start plus end frame on Wan 3.0 and describe the transition; this is the only Wan route on Makeartly with an end-frame slot.
You need a silent 5-second loop as cheaply as possible
Skip Wan 3.0 and use Wan 2.2 Spicy image-to-video; it finished 98% of runs with no review stops and a far lower quote.
- Create or upload a sharp, face-visible still of a fictional adult character, cropped to your target ratio.
- Open Wan 3.0 image-to-video in Studio and drop the still into the start-frame slot; add an end frame if you want a defined finishing pose.
- Write a motion-first prompt and leave Prompt enhance on for the first run.
- Render 3 seconds at 480P to check the motion, then re-run at 720P or 1080P for the final.
- If the review blocks the rendered clip, run the same still on Wan 2.2 Spicy instead of rewording.
Limitations and decision checks
Acceptance test the still before the clip
Treat the 3-second 480P render as the acceptance test: compare it frame by frame with the start still, stop if the face or hands have changed, and choose a cleaner still before spending on a longer tier.
Most image-to-video failures in our logs trace back to the input image, not to the prompt. End frames only work within one scene: If the end frame changes lighting, outfit or camera position, Wan 3.0 will invent a cut or morph between the two stills. Keep both frames in the same setup.
