Wan 2.2 TI2V-5B
A 5B-parameter dense model that handles both text-to-video and image-to-video in one checkpoint.
- Up to
- 5 seconds
- Resolution
- 1280x704 (720P)
- Weights
- Downloadable
- API
- Yes
- Commercial use
- Vendor states commercial use is permitted
It uses the Wan2.2-VAE with a 16x16x4 compression ratio to generate 720P at 24 fps on a single 24GB consumer GPU. This is the variant most people actually run locally.
Apache-2.0. No revenue threshold, MAU cap or territorial restriction.
Specification
- What it takes in
- text, image
- What it makes
- video
- Maximum duration
- 5 seconds
- Maximum resolution
- 1280x704 (720P)
- Aspect ratios
- not established
- Controls
- text prompt, optional reference image, negative prompt, seed, resolution selection, prompt extension, offloading and quantization flags
- Weights
- Downloadable
- API
- Yes
- Pricing
- Free open weights (Apache-2.0). Official README: 'at least 24GB VRAM (e.g. RTX 4090 GPU)' for single-GPU inference, making it the practical local option in the Wan 2.2 family. Hosted paid access via Alibaba Cloud Model Studio.
- Released or updated
- 2025-07-28
Supported aspect ratios is not established here. That means no page we could retrieve stated it, not that the model lacks one.
Strengths
- Runs on a single RTX 4090 (24GB)
- 720P at 24 fps natively, better temporal smoothness than 16 fps peers
- One checkpoint covers both T2V and I2V
- Apache-2.0
Weaknesses
- Noticeably weaker prompt adherence and motion than the A14B models
- 5B capacity limits complex multi-subject scenes
- No audio
- 720P ceiling
Best for
- Local and consumer-GPU video generation, prototyping, and cost-sensitive batch generation.
Where the sources disagree
Recorded rather than resolved. Picking the tidier number would hide the fact that the vendor's own pages do not agree.
- Some third-party guides state 121 frames as the maximum for TI2V-5B; the official README states 720P/24fps and references a 5-second clip but does not publish a hard frame cap.
What to understand before using it
Explainers naming the mechanisms this model's own documentation mentions.
MCP servers that front Alibaba Tongyi Lab (Wan)
Servers in the media MCP directory whose own description or homepage names this vendor. Absence of a server here is not evidence that none exists: most publish no tool list at all.
-
370.ai, AI Gateway: Video (Seedance 2.0, Wan, HappyHorse), Image, Speech + 100+ Chat Models
One key, 100+ models, chat with any LLM and generate video, images, speech. Free trial at 370.ai.
-
Aetherwave
Music, image, video and audio generation across top AI providers - one key, one credit pool.
-
Clipia
Generate AI images, video, speech, music and presentations from Claude, ChatGPT and Cursor.
-
Fal
Create and edit high-quality media across various formats including images, videos, audio, and 3D models.
-
Filtrix AI MCP
Filtrix MCP for image/video generation. Portal.
-
Higgsfield
Create images and videos from a prompt inside any MCP compatible AI tool.
See it from the vendor
Shotcraft hosts no sample output. Judging a model means watching its own showreel with your own eyes, and a clip re-hosted here would be a clip we chose.
Nearby in the directory
Wan 2.1 T2V (1.3B and 14B)
The first widely adopted Apache-2.0 video generation family. T2V-1.3B generates 480P and fits in about 8.19GB of VRAM; T2V-14B handles 480P and 720P.
Video generation
Wan 2.2 I2V-A14B
Animates a still image into video, optionally guided by a text prompt, using the same 27B/14B-active mixture-of-experts architecture as the text-to-vi
Video generation
Wan 2.2 S2V-14B
Audio-driven video generation: takes a reference image plus an audio track and produces a speaking or performing subject synchronized to that audio. O
Video generation