WAN 2.1 Image To Video 720P Ultra-Fast With LoRA Support

Wan 2.1 I2V 720p LoRA Ultra Fast — wavespeed-ai/wan-2.1/i2v-720p-lora-ultra-fast

Wan 2.1 I2V 720p LoRA Ultra Fast is a high-speed image-to-video model that turns a single reference image into a short, coherent clip guided by your prompt. It supports up to 3 LoRAs per run, making it great for fast style exploration, character looks, and branded aesthetics while keeping generation time low.

Key capabilities

• SOTA Performance: Wan2.1 consistently outperforms existing open-source models and state-of-the-art commercial solutions across multiple benchmarks.
• Multiple Tasks: Wan2.1 excels in Text-to-Video, Image-to-Video, Video Editing, Text-to-Image, and Video-to-Audio, advancing the field of video generation.
• Visual Text Generation: Wan2.1 is the first video model capable of generating both Chinese and English text, featuring robust text generation that enhances its practical applications.
• Powerful Video VAE: Wan-VAE delivers exceptional efficiency and performance, encoding and decoding 1080P videos of any length while preserving temporal information, making it an ideal foundation for video and image generation.

Use cases

• Bring key art to life: animate posters, portraits, product shots into short clips
• Style locking for content pipelines: apply one or more LoRAs for consistent aesthetics
• Character look development: combine identity/style LoRAs for repeatable outputs
• Social media loops: quick 5–10s clips from a single still image
• Marketing variations: same composition, different mood, lighting, wardrobe, brand style

Pricing

Inputs

• image (required): reference image that anchors subject and composition
• prompt (required): what happens in the clip (action + camera + mood)
• loras (optional): up to 3 LoRA items (each with a path and a scale)

Parameters

• duration: clip length (commonly 5s or 10s)
• num_inference_steps: more steps can improve motion coherence and detail
• guidance_scale: higher values follow the prompt more strongly
• strength: how strongly the model follows the input image vs. deviating into new variations
• negative_prompt: describe what to avoid (blur, artifacts, jitter, low quality)
• flow_shift: adjust motion behavior (useful for smoother or more dynamic movement)
• seed: set for reproducible results (-1 for random)
• loras[].path: LoRA identifier or a public safetensors URL
• loras[].scale: LoRA weight (tune per LoRA; start around 0.8–1.0)

Prompting guide (I2V + LoRA)

Write prompts like a director’s brief:

• Subject: who/what is on screen
• Action: what changes over time (gesture, expression, environment)
• Camera: push-in, pull-back, pan, tilt, handheld vs. locked-off
• Mood/lighting: golden hour, neon rain, candlelight, fog, rim light
• Motion style: subtle, smooth, energetic, dramatic

LoRA stacking tip: keep the base prompt clean, and let LoRAs handle style/identity. If multiple LoRAs conflict, lower one LoRA scale rather than overloading the prompt.

Example prompts

• A stylish street cat wearing sunglasses holds a skateboard. The cat shifts its stance, tail sways, and the camera slowly pushes in. Cinematic lighting, crisp detail, smooth natural motion, 720p, 5 seconds.
• A fantasy portrait comes alive: soft wind moves hair and fabric, embers drift across the frame, the character slowly turns their head and smiles. Slow dolly-in, moody lighting, subtle fog, smooth motion.