
Wan 2.6 vs Wan 2.7: What's New, What's Different, and When to Upgrade

TL;DR
| Wan 2.6 | Wan 2.7 | |
|---|---|---|
| Text-to-video | ✅ Multi-shot, 2-15s | ✅ Multi-shot, 2-15s (same) |
| Image-to-video | First frame only | First AND last frame + video continuation |
| Reference video | Single reference, no voice | Up to 5 refs + voice cloning |
| Video editing | ❌ Not available | ✅ New: instruction-based editing |
| Flash variants | ✅ Faster i2v and r2v | ❌ Not yet available |
| EvoLink pricing | Per-second rate | $0.086/sec at 720p |
1. What Wan 2.7 adds that Wan 2.6 does not have
Video editing (wan2.7-video-edit)
This is the single biggest difference. Wan 2.7 can take an existing video and apply text-guided edits — style transfer, background replacement, clothing changes, colorization — without re-generating from scratch.
Before Wan 2.7, editing a generated clip meant re-generating the entire thing with a modified prompt and hoping the output stayed close. Now you can pass the original clip plus an instruction like "change the background to a snowy mountain" and get an edited version that preserves the original motion.
{
"model": "wan2.7-video-edit",
"prompt": "Convert the scene to anime style",
"video_urls": ["https://your-cdn.com/source.mp4"],
"keep_original_sound": true,
"duration": 0
}Wan 2.6 has no equivalent. If your workflow involves post-generation iteration, this alone may justify the upgrade.
First-and-last-frame control (image-to-video)
image_start). You define where the video starts, and the model decides where it ends.image_end), so you can define both endpoints. The model infers the motion trajectory between them. This means:- Product rotation videos: define front view and back view, model fills the spin
- Scene transitions: define the opening and closing frame of a transition
- Looping content: set the last frame identical to the first for seamless loops
generation_mode:| Mode | Inputs | Wan 2.6 | Wan 2.7 |
|---|---|---|---|
first_frame | image_start | ✅ (implicit) | ✅ |
first_last_frame | image_start + image_end | ❌ | ✅ |
video_continuation | video_urls[0] + optional image_end | ❌ | ✅ |
Multi-character reference video with voice cloning
wan2.6-r2v) supports referencing a single video to carry character appearance into a new scene. No voice cloning.wan2.7-reference-video) supports:- Up to 5 reference inputs (images + videos combined)
- Voice cloning via
model_params.voice_bindings— map each character to a 1-10 second audio sample - Multi-character scenes — reference
Image 1,Image 2,Video 1in the prompt to place multiple characters
This turns reference video from a single-character consistency tool into a full multi-character production pipeline.
2. What Wan 2.6 still has that Wan 2.7 does not
Flash variants
wan2.6-image-to-video-flash, wan2.6-reference-video-flash) that trade some quality for faster generation and lower cost. These are ideal for:- A/B testing variants before committing to a final render
- Rapid iteration during concept development
- Volume workflows where speed matters more than peak quality
Proven production stability
Wan 2.6 has been in production on EvoLink since early 2026. Teams have established patterns around its behavior, duration limits, and Flash pricing. Wan 2.7 is newer, and while it passes the same endpoint and billing patterns, early-stage quirks are always possible.
3. Side-by-side comparison table
| Feature | Wan 2.6 | Wan 2.7 |
|---|---|---|
| Text-to-video | ✅ 2-15s, multi-shot | ✅ 2-15s, multi-shot |
| Image-to-video | First frame only | First + last frame, video continuation |
| Reference video | Single ref, no voice | Up to 5 refs, voice cloning |
| Video editing | ❌ | ✅ Instruction-based |
| Flash variants | ✅ (i2v, r2v) | ❌ |
| Max prompt length | 5000 chars | 5000 chars |
| Negative prompts | ✅ | ✅ |
| Audio handling | Auto-generated or driving audio | Same + voice cloning in R2V |
| Output quality | 720p / 1080p | 720p / 1080p |
| Aspect ratios | 16:9, 9:16, 1:1, 4:3, 3:4 | 16:9, 9:16, 1:1, 4:3, 3:4 |
| EvoLink pricing | Per-second rate | $0.086/sec (720p) |
| API endpoint | /v1/videos/generations | Same |
| Async pattern | Task ID + polling/callback | Same |
4. Decision tree: should you upgrade?
Do you need video editing?
├── Yes → Use Wan 2.7 (the only Wan route on EvoLink with this capability)
└── No
├── Do you need voice cloning or multi-character reference video?
│ ├── Yes → Use Wan 2.7
│ └── No
│ ├── Do you need first-and-last-frame control in I2V?
│ │ ├── Yes → Use Wan 2.7
│ │ └── No
│ │ ├── Do you need Flash variants for rapid iteration?
│ │ │ ├── Yes → Stay on Wan 2.6
│ │ │ └── No → Either works, Wan 2.7 is the newer default
│ │ └──
│ └──
└──
5. Migration: how to move from Wan 2.6 to Wan 2.7
What stays the same
- API endpoint:
POST /v1/videos/generations - Authentication: same API key, same Bearer token
- Async pattern: same task ID, same polling, same callback
- Billing: same EvoLink account and credit system
What you change
model parameter:| Wan 2.6 | Wan 2.7 |
|---|---|
wan2.6-text-to-video | wan2.7-text-to-video |
wan2.6-image-to-video | wan2.7-image-to-video |
wan2.6-reference-video | wan2.7-reference-video |
| — | wan2.7-video-edit (new) |
Your existing prompts work without modification. The upgrade is literally a string replacement in your model parameter.
Gradual migration strategy
- Route new projects to Wan 2.7 by default.
- Keep existing Wan 2.6 integrations running — no need to break what works.
- Migrate specific workflows when you need 2.7-only features (editing, voice, frame control).
- Keep Wan 2.6 Flash for workflows where iteration speed matters.
6. Pricing comparison
| Wan 2.6 Standard | Wan 2.6 Flash | Wan 2.7 | |
|---|---|---|---|
| 720p per second | Standard rate | Lower than standard | $0.086 |
| 1080p per second | 1.67× of 720p | 1.67× of Flash 720p | $0.144 |
| R2V billing | Input + output duration | Input + output duration | Input + output duration |
| Video edit billing | N/A | N/A | Input + output duration |
| Failed task billing | — | — | Free (R2V + Edit) |
7. FAQ
Is Wan 2.7 always better than Wan 2.6?
No. If you need Flash variants for rapid iteration, Wan 2.6 Flash is still faster and cheaper per iteration cycle. Both versions coexist on EvoLink for a reason.
Can I use both Wan 2.6 and Wan 2.7 on the same EvoLink account?
wan2.6-text-to-video and wan2.7-text-to-video in the same integration.Will Wan 2.6 be deprecated?
Not currently. Wan 2.6 and Wan 2.7 run in parallel on EvoLink. Wan 2.6 remains actively supported, especially for Flash variant workflows.
Do I need to rewrite my prompts for Wan 2.7?
No. Wan 2.7 accepts the same prompt format as Wan 2.6, including multi-shot narrative syntax. Your existing prompts work without modification.
What about Wan 2.5? Where does that fit?
Next steps
- Full integration guide: Wan 2.7 API guide
- Try the playground: Wan 2.7 model page | Wan 2.6 model page
- Family overview: Wan API family collection
- Pricing details: Wan API pricing guide


