First & Last Frame Control
Give the model an opening and a closing frame and let it solve the motion between them, so a shot starts and ends exactly where your edit needs it.
Open the generatorWan 2.7 AI Video Generator | First & Last Frame · Video Extension · Native Audio
Wan 2.7 is the newest video model from Alibaba's Tongyi Lab: first-and-last-frame control, video continuation, audio-driven performance, 720P/1080P output, and any duration from 2 to 15 seconds.
Wan 2.7 Video Highlights
From endpoint control to native audio, Wan 2.7 covers the controls that decide whether generated footage survives the edit.
Video Continuation
Extend an existing clip past its cut point while keeping subject, lighting, and camera behaviour consistent — useful when a take runs short.
See how it worksNative Audio & Voice Sync
Generate dialogue, ambience, and music together with the picture, or drive the performance from your own WAV/MP3 track of 2 to 30 seconds.
Generate with audioDuration From 2 to 15 Seconds
Set any whole-second length between 2 and 15 instead of choosing from a few fixed presets, so clips match the cut you already have.
Set a length720P & 1080P Across Five Ratios
Render natively at 16:9, 9:16, 1:1, 4:3, or 3:4 in either resolution tier, so vertical and square deliverables skip the reframe pass.
Pick a formatWan 2.7 Control Highlights
The Wan 2.7 upgrade is about direction, not just fidelity — endpoint control, clip extension, and audio timing give you handles on the parts of a shot that normally can't be steered.
Lock both ends of a shot
First-and-last-frame generation fixes the opening and closing frame, so a shot enters and exits exactly where your edit needs it.
Extend an existing clip
Video continuation grows a clip you already have past its cut point, keeping subject, lighting, and camera behaviour consistent.
Audio-driven performance
Attach a WAV or MP3 track between 2 and 30 seconds and let the voice or music drive timing, mouth movement, and expression.
720P and 1080P, five ratios
16:9, 9:16, 1:1, 4:3, and 3:4 are all native, so vertical and square deliverables don't need a separate reframe pass.
Duration to the second
Any whole number from 2 to 15 seconds instead of a few fixed presets — shots can match the cut you already have.
Prompt extension & negative prompts
Automatic prompt expansion fills in cinematic detail, while negative prompts hold back the artifacts you keep seeing.
What is Wan 2.7?
Wan 2.7 is the latest generation of the Wan video family from Alibaba's Tongyi Lab, released in April 2026. Built on a mixture-of-experts architecture, it generates video from text, images, reference material, and audio, with clearly better motion stability and detail retention than Wan 2.6.
The biggest change is endpoint control. Wan 2.7 image-to-video covers three tasks in one model: first-frame-to-video, first-and-last-frame-to-video, and video continuation. Locking both ends of a shot — or extending a clip you already have — turns generation from a lottery into something you can actually direct.
Audio is native rather than bolted on. Supply a driving track (WAV or MP3, 2 to 30 seconds) to drive a performance, or let the model produce dialogue, ambience, and music in sync with the picture. Output runs at 720P or 1080P across five aspect ratios, at any whole-second length from 2 to 15.
Why Wan 2.7?
Wan 2.7 puts control where production actually needs it: where a shot starts, where it ends, how long it runs, and what it sounds like.
🎨
First-and-last-frame control
Fix the opening and closing frame and let the model solve the motion between them — the most direct way to make generated footage fit an existing edit.
⚡
Video continuation
Feed in a clip and extend it forward, keeping subject, lighting, and camera behaviour consistent past the original cut point.
✨
Native audio & voice sync
Drive generation with an audio track, or generate dialogue, ambience, and music together with the picture so sound and image line up without a separate pass.
📱
1080P output, 2–15s duration
Pick 720P or 1080P across five aspect ratios, at any whole-second length from 2 to 15 — no rounding shots up to a fixed preset.
Create with Wan 2.7 in 3 steps
Choose your control points, describe the shot in camera language, then set length and sound.
1
Pick a mode
Start from text, from a single first frame, from a first-and-last frame pair, or from a clip you want to extend.
2
Write a strong prompt
Name the subject, the scene, and the camera language — shot size, movement, lens feel, lighting. Add a negative prompt for artifacts you keep seeing.
3
Set duration and audio
Choose resolution, aspect ratio, and an exact length in seconds. Attach a driving audio track if the performance needs to follow a voice or a beat.
Wan 2.7 Use Cases
How teams across industries put Wan 2.7's endpoint control, clip extension, and native audio to work on real deliverables.
Film & Production
Brand Marketing
Education & Training
Social Media
For film, shorts, and commercial work: endpoint control and clip extension make generated footage something an editor can actually cut with.For brand and e-commerce teams: more variants per brief, consistent brand expression, and faster localized creative.For course and training teams: turn concepts into camera-ready scenes and keep narration aligned without a dubbing pass.For creators: stronger openings, natural sound, and exact-length cuts that fit each platform's format.
Shots that fit the cut
Lock the first and last frame so a generated shot enters and exits exactly where the timeline needs it.
Extend a short take
Continue an existing clip past its cut point when a take lands two seconds short of the edit.
Dialogue & score in sync
Generate speech, ambience, and music alongside the picture to cut down on post alignment work.
1080P deliverables
Render at 1080P across five aspect ratios so previz and social cutdowns come out of the same pass.
FAQ
What is Wan 2.7?
Wan 2.7 is the latest video generation model in Alibaba Tongyi Lab's Wan family, released in April 2026. It generates video from text, images, reference material, and audio, and adds first-and-last-frame control plus video continuation on top of Wan 2.6's capabilities.
What does first-and-last-frame control do?
You supply both the opening and the closing frame of a shot, and the model generates the motion between them. It is the most reliable way to make a generated shot fit an edit, because both endpoints are fixed rather than left to the model.
How does video continuation work?
You give Wan 2.7 an existing clip and it extends the footage forward from the end of that clip, keeping the subject, lighting, and camera behaviour consistent. It is useful when a take comes up short of the length your timeline needs.
What resolutions and aspect ratios are supported?
Wan 2.7 offers a 720P and a 1080P tier, with 1080P as the default. Both tiers support 16:9, 9:16, 1:1, 4:3, and 3:4, so horizontal, vertical, and square deliverables are all native.
How long can a Wan 2.7 video be?
Any whole number of seconds from 2 to 15, with 5 seconds as the default. Unlike fixed presets, this lets you match an exact slot in an existing edit.
How does audio work in Wan 2.7?
Audio is generated together with the picture rather than added afterwards. You can also supply your own driving audio — WAV or MP3, between 2 and 30 seconds — to drive a performance so delivery and timing follow your recording.
Any tips for better Wan 2.7 prompts?
Be explicit about the subject, the scene, and the camera language: shot size, camera movement, lens feel, and lighting. Use a negative prompt for artifacts you keep seeing, and let prompt extension fill in cinematic detail when your prompt is short.
Which models does the generator on this page run?
The generator on this page currently runs Wan 2.6 and Wan 2.5 through the official DashScope API. Wan 2.7 API access is still being wired up and will appear in the model list here once it is live. Everything described on this page refers to Wan 2.7's own published specifications.
Ready to create with Wan 2.7?
Endpoint control, clip extension, native audio, and exact-length output — the controls that decide whether generated footage makes it into the edit.