Short answer: choose Seedance 2 when the creative plan needs several kinds of reference media, first-and-last-frame control, or a 4–15-second generation in up to seven aspect-ratio modes. Choose Wan 2.7 when negative prompts, prompt extension, 1080p output, or a connected reference-to-video/video-edit step is important. That is an editorial workflow recommendation drawn from current AI Image Video controls, not a quality ranking or a report of live testing.
For a shared generation comparison, start in Text to Video or Image to Video. Use Video to Video only when the real task is transformation or editing rather than creating the baseline clip.
Seedance 2 vs Wan 2.7 at a glance
| Decision factor | Seedance 2 | Wan 2.7 |
|---|---|---|
| Core generation workflows | Text, image, and reference to video | Text and image to video |
| Additional platform workflows | Reference package can include images, video, and audio | Separate reference-to-video and video-edit tools are available |
| Generation duration | 4–15 seconds | 2–15 seconds in the generation workspaces |
| Current resolution choices | 480p, 720p | 720p, 1080p |
| Aspect ratios | 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, Adaptive | 16:9, 9:16, 1:1, 4:3, 3:4 |
| Notable controls | First/last frame, reference images/video/audio, generated audio, web search | First/last frame, audio input, negative prompt, prompt extension, watermark control |
| Current credit unit | Per billable second | Per output second for generation; per billable second for reference/edit tools |
| Best workflow fit (editorial) | Reference-rich generation and endpoint framing | Prompt refinement plus a generation-to-edit pipeline |
Two different ways to control a shot
Seedance 2 concentrates control in the input package. A prompt can be combined with a first frame, a first and last frame, or reference images, reference videos, and reference audio. The current integration keeps frame-based and reference-media branches separate: first/last frames cannot be combined with reference media. It exposes 480p and 720p output from 4 to 15 seconds.
Wan 2.7 concentrates more control in text and downstream workflow. The generation integration lists first/last frame, a negative prompt, automatic prompt extension, watermark control, and optional audio input. It exposes 720p and 1080p and allows 2–15-second output in the generation workspaces.
Wan also has dedicated reference-to-video and video-edit contracts. Those tools currently accept 2–10-second jobs at 720p or 1080p. Reference-to-video can use up to five reference media items across images and videos, plus an optional first frame and reference voice. Video edit accepts a source video and can use a reference image. These are separate operations, so do not assume their inputs or duration rules apply to ordinary text-to-video generation.
Credit comparison on AI Image Video
The numbers below are platform credits, not provider pricing. They can change; check the live workspace estimate before submission.
Seedance 2 charges per billable second:
- 480p without reference video: 29 credits per output second.
- 720p without reference video: 62 credits per output second.
- 480p with reference video: 18 credits per billable second.
- 720p with reference video: 38 credits per billable second.
- With reference video, verified reference input time and generated output time are both billable.
A 5-second 720p output without reference video is 310 credits at the current schedule.
Wan 2.7 generation charges per output second:
- 720p: 24 credits per output second.
- 1080p: 36 credits per output second.
A 5-second generation is therefore 120 credits at 720p or 180 at 1080p. Wan 2.7 reference-to-video and video edit use the same current resolution rates but bill per billable second; source-duration rules can affect an edit charge. Treat generation, reference, and edit estimates as different line items.
Who should choose Seedance 2?
Seedance 2 is the more suitable editorial choice when:
- Reference video or reference audio belongs in the initial generation brief.
- The opening and ending composition need separate supplied frames.
- Adaptive, 21:9, or another Seedance-only ratio is needed.
- The team wants one reference-led generation stage rather than a later edit stage.
- The reference-video duration can be verified for credit approval.
Seedance’s wider input contract does not mean every job needs every asset. Decide whether the task is frame-led or reference-led, then upload only evidence that supports that branch.
Who should choose Wan 2.7?
Wan 2.7 is the more suitable editorial choice when:
- A negative prompt is important for describing exclusions.
- Prompt extension should be tested as an explicit switch.
- 1080p is required in the current workspace.
- The planned process includes a separate reference-to-video or video-edit pass.
- A source video’s original audio should be handled through the edit workflow.
For editing, start in Video to Video and record the source duration, requested duration behavior, resolution, audio setting, and displayed estimate. An editing result should not be compared to a fresh text-to-video result as though they were the same task.
A same-brief evaluation with a controlled second round
The first round should test only shared generation capabilities:
- Choose 5 seconds, 720p, and 16:9 in Text to Video.
- Turn Wan prompt extension off, leave its negative prompt blank, and keep Seedance web search off.
- Use the same prompt and match the audio choice. If an audio-input comparison is required, run that as a separate round.
- Generate an equal number of candidates and record the credits displayed before each request.
- Score required subject attributes, action completion, composition, camera instruction, background exclusions, start/end usability, and audio relevance.
- Repeat in Image to Video using the same starting still and motion instruction.
Example prompt: “Five-second 16:9 studio shot of a folded yellow rain jacket on a dark wooden bench. Camera pushes in slowly as one sleeve lifts in a light breeze, soft overcast window light, one continuous shot, no people, logos, labels, or on-screen text.”
For the second round, test the capability that motivates the choice. Add a necessary reference video or audio file to Seedance, or add a precise negative prompt and toggle prompt extension for Wan. Change one variable per generation. If the deliverable is an edit, generate or provide one baseline source, then evaluate Wan video edit against written transformation criteria rather than against a different source clip.
Limitations
- This page documents current product capabilities and credit rules; it does not claim live benchmark results, measured speed, or superior visual quality.
- Wan generation and Wan reference/edit tools have different maximum durations and billing semantics.
- Seedance’s reference-video rate applies to billable input plus output time, not output alone.
- A negative prompt or additional reference is a control, not a guarantee that every unwanted element will be absent.
- Generated results vary. Use multiple controlled candidates for a project decision without turning that sample into a global model ranking.
- Confirm rights and consent for every image, video, voice, or likeness used as input.
Related comparisons
- Seedance 2 vs Kling 3: reference inputs and mode selection
- Seedance 2 vs Veo 3.1: flexible duration versus fixed variants
- Seedance 2 vs HappyHorse: mixed-media versus image-reference planning
- Seedance 2 vs Sora 2
FAQs
Is Wan 2.7 cheaper than Seedance 2?
For some like-for-like platform settings, the arithmetic may produce a lower Wan estimate, but the billing units and supported workflows differ. Seedance reference-video jobs include input plus output billable time. Check the exact operation and displayed estimate instead of treating one rate as a universal price comparison.
Do both support image-to-video?
Yes. Both are available in the image-to-video workspace. Both also expose first-frame controls; their additional reference and prompt controls differ.
Which supports 1080p?
Wan 2.7 currently exposes 720p and 1080p. The current Seedance 2 site variants expose 480p and 720p.
Can Wan 2.7 edit an existing video?
Yes. A separate Wan 2.7 video-edit model is available in the video-to-video workflow. It is not the same operation as Wan text-to-video or image-to-video.
Where should I begin?
Use Text to Video for a clean prompt baseline, Image to Video for the same start-frame test, and Video to Video only if editing is part of the actual requirement.