Cosmos3-Super-Image2Video
Upload one start image, describe the motion you want, and Cosmos 3 Super renders a 1–7 second MP4 clip. Five aspect ratios, 10 credits per second of output.
Single-subject frames of the kind that work well as a clip's first frame — a clear subject, readable composition, and room for motion. Send one to the video workspace and describe the motion.
This image depicts a colossal blue dragon merging with mountainous terrain, featuring mineral-like scales and translucent wings. Light and shadow create a mystical atmosphere. Suitable for fantasy creature illustrations and film concept design visual references.
A cel-shaded forest elf ranger standing among tall trees, rendered with clean outlines and muted green tones. Suitable for fantasy character design and game concept art projects.
A man in casual attire poses on a luxury yacht staircase, framed by a suspension bridge and coastal city. Warm golden sunlight creates distinct shadows and highlights. Suitable for high-end lifestyle photography and commercial advertising image projects.
A photorealistic portrait of a red-haired woman holding a white rabbit in a golden meadow, featuring natural lighting and authentic textures. The scene evokes a serene, warm countryside atmosphere. Suitable for nature-themed illustrations and character poster design.
A magenta-hued cat sits within a dense sea of blue flowers, creating a striking complementary color contrast. Suitable for pet illustration and dreamlike nature art projects.
This image prompt evokes a nostalgic summer afternoon at a suburban train station, featuring a Japanese high school girl in a sailor uniform bathed in soft golden light. The scene captures gentle wind and reflective glass, creating a cinematic, slice-of-life atmosphere. Suitable for character portrait design and story illustration projects.
These images come from our prompt library and were not all made with Cosmos 3 Super. Your results will look different.
The box above hands your image, prompt, and settings to the video workspace — nothing is generated until you press the button there. If you do not have a start image yet, generate one with the text-to-image tool on the homepage and come back.
One image is required and it becomes the clip's first frame. A sharp, well-composed source gives the model the most to work with.
Write what should move and how the camera should behave — a slow orbit, a gentle push-in, drifting clouds. The model animates the scene rather than redrawing it from scratch, so concrete motion verbs beat style adjectives here.
Pick one of five aspect ratios and a length between 1 and 7 seconds. The start image is resized and center-cropped to your chosen ratio, so compose with that crop in mind.
Each second of output costs 10 credits. The finished MP4 lands in your history, where you can replay it and download the file.
What the image-to-video variant accepts and what it returns, exactly as the dashboard exposes it.
| Required input | One start image plus a prompt (text-only is not supported) |
|---|---|
| Aspect ratios | 16:9, 4:3, 1:1, 3:4, 9:16 — start image is resized and center-cropped |
| Duration | 1–7 seconds |
| Cost | 10 credits per second of output |
| Output format | MP4 video file |
| Audio | None — the output has no audio track |
Video is billed per second of output — 10 credits for every second the clip runs.
| Duration | Cost |
|---|---|
| 1 s | 10 credits |
| 2 s | 20 credits |
| 3 s | 30 credits |
| 4 s | 40 credits |
| 5 s | 50 credits |
| 6 s | 60 credits |
| 7 s | 70 credits |
The prompt describes what moves and how — not a new scene. A few patterns work consistently.
Name the move plainly: slow orbit, gentle push-in, pull-back, pan left, static shot. One camera move per clip reads better than several stacked together.
Say what the subject does — turns, walks, waves, spins — and roughly how fast. “Slowly turns its head toward the camera” works better than “cinematic movement”.
Weight, wind, and liquid read as physical detail: curtains stirring, steam rising, hair blowing, water rippling, a flag snapping taut.
The dashboard’s negative prompt field can exclude artifacts — warping, flicker, morphing faces, extra limbs — instead of fixing them afterward.
The start image becomes the clip’s first frame, so it deserves a minute of care.
The uploader accepts common image formats — JPEG, PNG, and WebP all work.
The image is resized and center-cropped to your chosen ratio. Keep what matters near the middle, or pick the ratio that matches the source.
A single readable subject — one product, one face, one landmark — animates more predictably than a crowded frame.
Light already in the scene guides the motion the model adds. One clear light direction gives it more to work with than flat or mixed lighting.
The jobs a short, silent clip actually does well: adding motion to a still, giving depth a second of drift, and turning a single frame into a feed-ready format.
A static packshot becomes a slow orbit or a light sweep. Useful for product pages and ad units that need motion without a reshoot.
Depth in a landscape photo turns into a gentle drift across the frame. Works well for hero backgrounds and travel content, where a few seconds of drift keep a page from feeling static.
Machinery, workshops, and hardware gain ambient motion — spinning fans, moving light, drifting dust — for demos and documentation.
The 9:16 ratio turns a phone-ready image into a feed-length vertical clip without separate editing.
These images come from our prompt library and were not all made with Cosmos 3 Super. Your results will look different.
Read these before you spend credits — they describe what the variant can and cannot do on this site, not just what it costs.
There is no text-only video mode. Every run needs one start image and a prompt.
The start image is resized and center-cropped to the chosen ratio. Keep the subject near the center if the composition matters.
Clips run 1–7 seconds. Longer sequences need to be built from several clips.
If a task fails, the credits it reserved are refunded automatically.
10 credits per second of output — a 3-second clip is 30 credits and the maximum 7 seconds is 70. Credits for failed tasks are refunded.
Yes. The start image can be any image you upload — a photo, a render, or an image you generated earlier. One image per run.
No. The output is a silent MP4 file.
It is resized and center-cropped to the aspect ratio you pick, then used as the clip's first frame.
No — a start image is required. If you need one first, generate it with the text-to-image tool on the homepage, then send the result here through the dashboard's video entry.
Upload a frame, describe the motion, and generate a clip in the video workspace.
