Image-to-video generator checklist for creators
Use this checklist before turning images into AI video clips: reference quality, motion direction, model fit, duration, and cost controls.
TL;DR
Before hitting generate on image-to-video, check 5 things: reference image quality, motion description clarity, model fit for your use case, expected duration, and StarPoints cost. Skipping any of these leads to wasted generations.
1. Prepare the reference image
Use a clear first frame with the subject, background, and desired composition already visible. Resolution should match your target output. A strong input image usually reduces unwanted motion artifacts. Avoid busy backgrounds — they confuse motion prediction.
2. Write motion instructions
Describe camera movement (pan, zoom, dolly), subject movement (walking, turning, gesturing), mood, and timing. Short, direct motion prompts are easier to reuse across multiple tests. Example: "Slow zoom-in on the character's face, soft lighting, cinematic depth of field, 5 seconds."
3. Choose the right model
Wan 2.7 (30 SP/5s, Pro): Best for general image-to-video with good quality-to-cost ratio. Seedance 2.0 (450 SP/5s, Pro): Best for story animation and complex motion. Start with Wan 2.7 for drafts, move to Seedance 2.0 for final output.
4. Set duration and aspect ratio
5 seconds is the standard for most models. 16:9 for widescreen/YouTube. 9:16 for vertical/shorts/TikTok. 1:1 for social media square posts. Match your output dimensions to the target platform before generating.
5. Check cost before generating
Wan 2.7: 30 SP/5s. Seedance 2.0: 450 SP/5s. Factor in 2-3 retries per scene. For a 5-scene project: budget 150-750 SP minimum for Wan 2.7, or 2,250-6,750 SP for Seedance 2.0.
Limitations
Image-to-video models struggle with: fine facial expressions during motion, fast action sequences, multi-subject scenes, and precise lip sync. Always generate test clips before committing to a full project. Expect 1-3 retries per scene for acceptable results.