Wan 2.7 video model suite now available on Together AI

Wan 2.7 video model suite now available on Together AI

⚡️ FlashAttention-4: up to 1.3× faster than cuDNN on NVIDIA Blackwell →

Introducing Together AI's new look →

🔎 ATLAS: runtime-learning accelerators delivering up to 4x faster LLM inference →

⚡ Together GPU Clusters: self-service NVIDIA GPUs, now generally available →

📦 Batch Inference API: Process billions of tokens at 50% lower cost for most models →

🪛 Fine-Tuning Platform Upgrades: Larger Models, Longer Contexts →

All blog posts

Model Library

Published 4/3/2026

Wan 2.7 video model suite now available on Together AI

A four-model video suite for generation, continuation, reference-driven workflows, and editing, rolling out on Together AI starting with text-to-video.

Authors

Arielle Fidel, Necoline Hubner, Zain Hasan, Derek Petersen, Blaine Kasten, Hiral Jasani, Sonny Khan

Table of contents

40+ Models Chosen for Production...40+ Models Chosen for Production...40+ Models Chosen for Production...

Links in this article

Wan 2.7 Model Page Wan 2.7 T2V in the Playground Wan 2.7 T2V Quickstart Video Generation Overview Contact Sales

Summary

Four-model suite: Wan 2.7 brings video generation, continuation, and editing to Together AI, starting today with text-to-video and expanding soon to image-to-video, reference-to-video, and video edit. Tighter creative control: Drive generation with optional audio, frame-level conditioning, reference inputs, and continuation workflows to reduce workflow fragmentation. Available on Together AI: Wan 2.7 runs on Together AI, the AI Native Cloud, with the same fast, reliable APIs, authentication, and billing surface developers already use across the rest of their multimodal stack. Simple pricing: Available starting at $0.10 per second of generated video through the Together AI API.

AI video is easy to generate and hard to steer. A team can get a promising clip from a prompt, but continuing it, matching a reference, or revising it without starting over usually means leaving the model that made it and patching the rest together somewhere else. The more control a project needs, the more the workflow turns into re-renders, handoffs, and manual cleanup. That is the gap Wan 2.7 is built to close across generation, continuation, reference-driven workflows, and editing. On Together AI, that expanded control surface becomes one platform instead of another disconnected toolchain. Wan 2.7 comes to Together AI, the AI Native Cloud, as a four-model suite rolling out from text-to-video into image-to-video, reference-to-video, and video edit. That gives teams a clearer path from first generation to continuation, reference-driven control, and revision through the same APIs, authentication, and billing surface they already use across the rest of their multimodal stack. Text-to-video available now Wan 2.7 Text-to-Video ( Wan-AI/wan2.7-t2v ) is available today on Together AI. It provides a stronger starting point for campaign content, product videos, and creative prototyping than a plain prompt-to-video surface by supporting: Flexible resolution: 720P and 1080P generation. Duration control: Video outputs ranging from 2 to 15 seconds. Audio support: Optional audio input to drive the generation. Prompt-driven direction: Multi-shot narrative control directly through prompt language.

WAN 2.7

5s

Your browser does not support the video tag.

Epic snowy train shot with cinematic scale.

$0.10/video

Try now

WAN 2.7

4s

Your browser does not support the video tag.

Sci-fi astronaut corridor with red emergency lighting.

$0.10/video

Try now

WAN 2.7

4s

Your browser does not support the video tag.

Luxury rooftop fashion film with flowing motion.

$0.10/video

Try now

WAN 2.7

5s

Your browser does not support the video tag.

Photoreal hummingbird macro with rainforest wonder.

$0.10/video

Try now

Image-to-video and reference-to-video coming soon As the rest of the Wan 2.7 suite rolls out, developers will get more control over how video is driven and structured. Image-to-Video ( Wan-AI/wan2.7-i2v ) supports: Advanced workflows: First-frame, first-and-last-frame, and continuation generation. Audio-video sync: Synchronized generation driven by audio inputs. Storyboarding: 3x3 grid-to-video generation workflows. Flexible outputs: 720P or 1080P generation up to 15 seconds.

Reference-to-Video ( Wan-AI/wan2.7-r2v ) supports: Reference inputs: Driven by reference image and reference video inputs. Complex scenes: Single-shot and multi-shot workflows, plus multi-character interactions. Outputs: 720P and 1080P generation up to 10 seconds.

Video edit coming soon Wan 2.7 Video Edit ( Wan-AI/wan2.7-edit ) gives teams a more direct way to modify footage without bouncing into separate editing systems for every pass. It extends the suite with: Instruction & reference editing: Modify footage via text instructions or reference image-based editing. Style transfer: Apply video style transfer to existing clips. Temporal feature transfer: Clone motion, camera work, effects, and style from source media.

Instead of splitting those jobs across separate tools, they stay inside one coordinated workflow, which reduces handoffs and makes iteration easier to manage. Try it now The Wan 2.7 Text-to-Video model is available today on Together AI Serverless Inference starting at $0.10 per second of generated video through the endpoint Wan-AI/wan2.7-t2v . If you are already using Together AI for text or image inference, adding video generation works the same way: Same authentication Same SDKs Same billing dashboard Transparent per-model pricing

Check out the Wan 2.7 Quickstart for full parameters (like audio inputs and resolution control), or copy the polling loop below to get started immediately:

import time from together import Together

client = Together()

job = client.videos.create( model="Wan-AI/wan2.7-t2v", prompt="A cinematic product video of a running shoe on wet pavement, smooth camera arc, dramatic reflections", resolution="1080P", ratio="16:9", seconds="5" )

print(f"Job ID: {job.id}")

poll until the video is completed

while True: status = client.videos.retrieve(job.id)

if status.status == "completed": print(f"Video URL: {status.outputs.video_url}") break elif status.status == "failed": print(f"Error: {status.error}") break

time.sleep(5)

)

Production deployment Start with serverless endpoints for development and testing. On Together AI, teams can move from experimentation to production on the same…