- ホーム
- Seedance AI 動画ジェネレーター
- 画像から動画
Seedance image to video
Image to video starts from a still you upload, a first frame and optionally a last frame, and generates the motion between. It is the workflow for animating product photos, real-estate stills, illustrations and storyboards while keeping the source exactly as it looks. Every active Seedance model on Seedancee supports it.
Animate an image
How Seedance image to video works
You provide one image as the first frame and, if you want to control where the shot ends, a second image as the last frame. The prompt then describes what happens in between and how the camera moves. Because the model is anchored to your pixels, identity, layout and colour stay far more consistent than with text alone. Seedance 2.0 and 2.5 can also return the final frame of a generation so you can use it as the first frame of the next shot and build sequences.
- 01
Upload a first frame
JPEG, PNG or WebP (2.x models also accept BMP, TIFF and GIF). Keep the aspect ratio close to your target output; the model can adapt, but cropping is cleaner.
- 02
Optionally add a last frame
Use it when the end state matters: a lid closed, a door open, a sunrise complete. Leave it out for open-ended motion.
- 03
Describe the motion
Say what moves, what stays still and how the camera behaves. "Furniture stays exactly as pictured" is a legitimate instruction.
- 04
Choose model and format
Match the resolution to the source; a 1080p output from a small phone screenshot will look soft.
Input requirements
Limits come from Kie.ai's documentation and are enforced in the generator from the model registry. For Seedance 2.0 and 2.5: JPEG, JPG, PNG, WebP, BMP, TIFF or GIF up to 30 MB, between 300 and 6000 pixels on each side, aspect ratio between 0.4 and 2.5. For Seedance 1.5 Pro: JPEG, JPG, PNG or WebP up to 10 MB, up to two images. Uploads are stored in your asset library so you can reuse a frame across generations.
| Seedance 2.5デフォルト | Seedance 2.0 | Seedance 2.0 Fast | Seedance 2.0 Mini | Seedance 1.5 Pro | |
|---|---|---|---|---|---|
| おすすめの用途 | Longer reference-driven scenes up to 30s with native audio and editing workflows. | High-quality 4–15s shots with up to 4K output and multimodal references. | Quicker iterations at 480p/720p with the same reference inputs as 2.0. | Lowest-cost drafts and social clips with strong visual consistency. | Previous-generation text and image-to-video with a fixed-lens option. |
| 長さ | 4–30s | 4–15s | 4–15s | 4–15s | 4–12s |
| 解像度 | 480p, 720p, 1080p | 480p, 720p, 1080p, 4k | 480p, 720p | 480p, 720p | 480p, 720p, 1080p |
| アスペクト比 | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 |
| 参照入力 | 開始/終了フレーム · 画像 30 枚 · 動画 10 本 · 音声 10 件 | 開始/終了フレーム · 画像 9 枚 · 動画 3 本 · 音声 3 件 | 開始/終了フレーム · 画像 9 枚 · 動画 3 本 · 音声 3 件 | 開始/終了フレーム · 画像 9 枚 · 動画 3 本 · 音声 3 件 | 開始/終了フレーム |
| ネイティブ音声 | あり | あり | あり | あり | あり |
| 編集/延長 | あり | なし | なし | なし | なし |
| 相対的な速度 | 標準 | 低速・高品質 | 高速 | 高速 | 標準 |
| 推定クレジット | 1 回の生成あたり 300〜5647 | 1 回の生成あたり 200〜3726 | 1 回の生成あたり 120〜450 | 1 回の生成あたり 40〜150 | 1 回の生成あたり 20〜257 |
| 機能の確認日 | 2026-09-29 | 2026-09-29 | 2026-09-29 | 2026-09-29 | 2026-09-29 |
Prompting for image to video
The image already carries the subject, setting and lighting, so the prompt should spend its words on motion and camera. Describe the change over time, name what must stay fixed, and give one camera instruction. Avoid re-describing the image in detail; conflicting descriptions cause the model to drift away from your source. Add the sound in its own final sentence as usual.
- Good: "Steam rises from the cup, the candle flickers, everything else stays still. Slow push-in."
- Weak: "A beautiful cozy cafe scene with a cup of coffee on a wooden table with warm light." (already in the image; no motion described)
- For first-to-last frame: describe the transition, not the two states. "The sky brightens from night to sunrise; streetlights fade as traffic starts."
Example image-to-video prompts
Starting from the uploaded product photo, a gentle quarter-turn orbit around the bottle as a soft light sweep crosses the label; background, colours and label stay exactly as pictured. Sound: quiet studio room tone.
- mode:
- image-to-video
- duration:
- 6s
- ratio:
- 1:1
- resolution:
- 1080p
- audio:
- on
From the uploaded living-room photo, a slow steady gimbal push toward the window as afternoon light drifts across the floor; no people, no camera shake, furniture unchanged. Sound: faint birdsong, quiet room.
- mode:
- image-to-video
- duration:
- 10s
- ratio:
- 16:9
- resolution:
- 1080p
- audio:
- on
Transition from the first frame (closed kraft box on a table) to the last frame (box open with the folded scarf visible): the lid lifts smoothly, tissue paper parts, the scarf catches the light. Top-down camera, locked. Sound: paper rustle, a soft thump.
- mode:
- image-to-video
- duration:
- 8s
- ratio:
- 4:3
- resolution:
- 1080p
- audio:
- on
Animate the uploaded watercolour illustration: rain falls across the street, the umbrella tilts slightly, puddles ripple, the painted texture stays visible throughout. Camera static. Sound: rain on pavement.
- mode:
- image-to-video
- duration:
- 6s
- ratio:
- 3:4
- resolution:
- 720p
- audio:
- on
Image to video vs reference to video
Image to video pins the first (and last) frame; the output begins as your image. Reference to video uses images, clips and audio as guidance for identity, style and motion without forcing the first frame to match. Use image-to-video when the shot must start from your exact picture; use reference-to-video when you need a character or product to appear consistently in a scene you describe from scratch. Seedance 2.5 accepts both kinds of input in one request.
Seedance reference to video
Multi-image, video and audio references on Seedance 2.0 and 2.5.
Cost
Image-to-video costs the same as text-to-video for a given model, duration, resolution and audio setting; input images carry no surcharge. Failed generations are refunded automatically. It works on both plans: Starter with Seedance 2.0 Mini, 2.0 Fast and 1.5 Pro at up to 720p and 10 seconds, Pro with every model at up to 1080p (4K on Seedance 2.0). The pricing page has the estimator and examples.
Pricing and credit estimator
よくある質問
Is Seedance image to video free?
Uploading and configuring is free without an account; generating needs a subscription and costs credits. There are no free trial credits; plans start at $4.99 a week for Starter.
What image formats does Seedance accept?
Seedance 2.0 and 2.5: JPEG, PNG, WebP, BMP, TIFF and GIF up to 30 MB, 300–6000 px per side. Seedance 1.5 Pro: JPEG, PNG and WebP up to 10 MB. These limits are enforced in the generator from the registry.
Can I set both a first and a last frame?
Yes, on every active model. Seedance 1.5 Pro is limited to those two images; 2.0 and 2.5 can add reference media on top.
Does image to video keep my subject consistent?
Far more than text-to-video, because the output starts from your pixels. For consistency across several shots, combine a first frame with reference images on Seedance 2.5 and reuse the returned last frame as the next first frame.
Does image to video have audio?
Yes. Native audio is generated from the sound description in your prompt on every active model.
Which model is best for image to video?
Seedance 2.0 for the sharpest short clips up to 4K, Seedance 2.5 for longer motion and returned last frames (both on the Pro plan), Seedance 2.0 Mini for cheap tests, and Seedance 1.5 Pro when the camera must stay locked (both also on Starter).