Alibaba · Open Source · Video Model
WAN 2.1 & WAN 2.7 — open-source AI video that actually converts
Alibaba's WAN models deliver natural-motion, physics-accurate video at 720p. Text-to-video, image-to-video, vertical format — all free to use, available on UGCad.ai for instant DTC ad creation.
On this page
Overview
What is WAN 2.1 / WAN 2.7?
WAN (Wanxiang) is Alibaba's open-source AI video generation model family. Developed by the Tongyi team, WAN 2.1 emerged in early 2025 as one of the most capable open-source video models available — and quickly became one of the most downloaded models on Hugging Face.
WAN 2.7 is the latest iteration, bringing improved motion coherence, better understanding of physical interactions, and stronger prompt fidelity. Both versions support text-to-video (T2V) and image-to-video (I2V), meaning you can either describe a scene or start from a product photo and animate it.
What makes WAN compelling for DTC brands is its strength in natural, everyday motion — a serum bottle tilting in light, a sneaker flexing on a surface, a supplement pouch opening. The outputs feel authentic rather than artificially cinematic, which tends to outperform on performance channels like TikTok and Meta.
WAN 2.1 & 2.7 are open-source models
available free via Hugging Face
available free via Hugging Face
Capabilities
What WAN 2.7 does best
Six standout capabilities that make WAN 2.1 and 2.7 valuable for DTC ad creation.
Natural physics simulation
WAN excels at realistic material behaviour — liquids pour, fabrics drape, surfaces reflect. Great for beauty, food, and wellness product shots where the physical interaction is the ad.
Image-to-video (I2V)
Start from your existing product photography. WAN animates still images into natural-motion video clips, turning your catalogue shots into scroll-stopping ads without reshooting.
Vertical 9:16 native support
WAN natively outputs in 9:16 vertical format — purpose-built for TikTok, Instagram Reels, and YouTube Shorts. No cropping or black bars required.
Fully open source
Both WAN 2.1 and WAN 2.7 are open-source and available on Hugging Face under a commercial-use-friendly licence. Self-host on your own GPU or run via UGCad.ai with no setup required.
Authentic UGC motion
Unlike cinematic models that over-produce movement, WAN's outputs feel hand-held and organic — the casual, authentic aesthetic that earns trust on performance ad channels.
Fast iteration cycles
WAN's lighter architecture relative to closed-source competitors enables rapid prompt iteration. Test 5–10 creative variations in the time it takes closed models to generate one.
Use cases
DTC ad formats WAN excels at
Three proven formats where WAN's natural-motion strengths translate directly into higher ad performance.
Product animation
Still photo → moving ad
Feed a product image into WAN's I2V pipeline and get a natural-motion video in under 2 minutes. Ideal for brands with strong photography but no video budget.
Lifestyle UGC
Authentic usage clips
WAN generates the organic, slightly imperfect motion that makes a clip feel user-shot rather than studio-produced. Perfect for skincare routines, supplement unboxings, and apparel try-ons.
Volume testing
Rapid creative variation
WAN's open-source speed advantage enables high-volume creative testing. Generate 20+ variations of an ad concept in hours and let performance data decide the winner.
Prompt guide
WAN prompts that perform for DTC
Proven prompt structures for WAN 2.1 and 2.7 across four DTC verticals. Copy, adapt, iterate.
Beauty / Skincare
"Close-up of a glass serum dropper held over a clean countertop, golden liquid falling in slow motion into a pool of water, natural window light, soft bokeh background, vertical 9:16"
💡 WAN handles liquid physics particularly well — lead with the physical interaction
Health / Supplements
"Hand reaching into frame and picking up a white supplement bottle from a marble surface, clean minimal kitchen background, warm morning light, casual hand-held feel, 9:16 vertical"
💡 Describe the hand motion explicitly — WAN's I2V mode is excellent for this
Food & Beverage
"Overhead shot of a cold-brew coffee being poured into a glass with ice, dark liquid mixing with cream, micro droplets rising, condensation on glass, bright natural light, square 1:1"
💡 Food & liquid scenes are a WAN strength — be specific about the substance behaviour
Apparel / Fashion
"White linen shirt on a person walking through a sunny doorway, fabric billowing slightly in breeze, shadows and light playing across cloth, lifestyle feel, soft shoulder camera movement, 9:16"
💡 WAN captures fabric drape and light interaction naturally — avoid over-specifying motion direction
Getting started
How to use WAN on UGCad.ai
No GPU, no setup, no waiting. WAN-powered video generation in four steps.
1
Choose WAN 2.7
Open UGCad.ai and select WAN 2.7 (or WAN 2.1) from the model selector. Pick your output format — vertical 9:16 for social or 16:9 for display.
2
Upload or describe
Either upload a product photo for image-to-video, or type a text prompt describing your desired scene. UGCad.ai's prompt assistant helps optimise for WAN's strengths.
3
Generate & iterate
WAN renders your clip in under 2 minutes on UGCad.ai's cloud infrastructure. Review the output and tweak your prompt — iterate rapidly until the motion is exactly right.
4
Export & launch
Download your WAN video and launch directly to TikTok, Meta, or YouTube. Add captions and hook text inside UGCad.ai before exporting for a fully production-ready ad.
Pricing
How much does WAN cost?
WAN is open-source and free to run locally. For most brands, a managed platform like UGCad.ai is the fastest and cheapest production path.
Self-hosted
Free to download
Run WAN 2.1 or 2.7 on your own GPU via Hugging Face. Requires technical setup and hardware costs.
- WAN 2.1 & 2.7 model weights
- Full model access, no limits
- Requires RTX 3090+ recommended
- Setup time: hours to days
Best for DTC brands
UGCad.ai
$29/month
WAN 2.7 plus every other leading video and image model, optimised for DTC ad formats. No GPU required.
- WAN 2.7 + WAN 2.1 included
- Kling, Veo 3, Runway, Seedance also included
- 9:16 vertical, 1:1 square, 16:9 horizontal
- Renders in under 2 minutes
- Commercial use licence included
Cloud API (Replicate / fal.ai)
~$0.05–0.15/video
Pay-per-generation via inference providers. Good for developers; not optimised for ad creation workflows.
- WAN 2.1 available on Replicate
- Per-second billing
- No ad-specific templates
- Developer setup required
Comparison
WAN 2.7 vs other video models
How WAN stacks up against leading AI video models for DTC ad creation.
| Model | Max resolution | I2V support | Open source | Natural motion | Best for |
|---|---|---|---|---|---|
| WAN 2.7 | 720p | ✓ | ✓ | ⭐⭐⭐⭐⭐ | Natural product motion, volume testing |
| Kling AI 2.0 | 1080p | ✓ | ✗ | ⭐⭐⭐⭐ | Cinematic UGC, premium brand feel |
| Veo 3 | 1080p | ✓ | ✗ | ⭐⭐⭐⭐ | Photorealistic video with audio |
| Runway ML Gen-3 | 1080p | ✓ | ✗ | ⭐⭐⭐ | Cinematic B-roll, brand lifestyle |
| Seedance 2.0 | 1080p | ✓ | ✗ | ⭐⭐⭐⭐ | High-throughput ad volume |
| Hailuo AI | 720p | ✓ | ✗ | ⭐⭐⭐⭐ | Emotional lifestyle, beauty/wellness |
FAQ
WAN 2.1 & 2.7 — common questions
WAN 2.1 is an open-source AI video generation model developed by Alibaba's Tongyi Wanxiang team. It supports text-to-video and image-to-video generation with realistic motion, natural physics simulation, and up to 720p output. It became one of the most downloaded open-source video models on Hugging Face in early 2025.
WAN 2.7 is the latest version of Alibaba's WAN model family. It improves on WAN 2.1 with better motion coherence, stronger prompt adherence, and enhanced detail rendering. Like WAN 2.1, it is open-source and commercially usable, available on UGCad.ai without any GPU setup.
WAN 2.1 and WAN 2.7 are open-source models released under permissive licences — free to self-host. For commercial use without managing GPU infrastructure, UGCad.ai offers WAN-powered video generation starting at $29/month, including ready-to-use DTC ad formats.
WAN 2.1 supports output resolutions up to 720p (1280×720) with aspect ratios including 16:9, 9:16 vertical, and 1:1 square. For most DTC ad use cases, 9:16 vertical at 720p is ideal for TikTok, Reels, and Shorts.
WAN excels at natural motion and physics realism for everyday product scenarios. Kling AI produces higher-resolution output (up to 1080p) and handles complex camera movements better. For cost-conscious brands or open-source deployments, WAN delivers strong quality at lower cost. UGCad.ai offers both — test which performs best for your product.
Yes. WAN 2.1 supports image-to-video (I2V) — you provide a product photo and the model animates it with realistic motion. This is particularly useful for DTC brands wanting to turn existing product photography into scroll-stopping video ads.
WAN's strength is natural, physics-accurate motion — liquids pour realistically, fabrics move naturally, hands interact convincingly with products. This produces UGC-style clips that feel authentic and hand-held rather than overly polished, matching the aesthetic that performs well on TikTok and Meta.
WAN 2.1 was developed by Alibaba's Tongyi Wanxiang team (Wan-AI) and released as an open-source model on Hugging Face. The model family is part of Alibaba's broader push into multimodal AI, competing with closed-source models from Runway, Kling, and Google.
WAN 2.7, like WAN 2.1, is released under a licence that permits commercial use. Always verify the specific terms on the model's Hugging Face repository. On UGCad.ai, all WAN-generated content is cleared for commercial use in your brand's ad campaigns.
On consumer GPUs (RTX 4090 class), a 5-second WAN 2.1 clip at 720p typically takes 3–8 minutes. On cloud infrastructure, the same render completes in 30–90 seconds. UGCad.ai runs WAN on optimised cloud hardware, so most clips are ready in under 2 minutes.
Generate WAN 2.7 videos today
No GPU, no setup. Open-source quality, production-ready speed — all on UGCad.ai.
Try WAN on UGCad.ai →