wavespeedai/wan-2.1-t2v-720p

OfficialView on Replicate →

Accelerated inference for Wan 2.1 14B text to video with high resolution, a comprehensive and open suite of video foundation models that pushes the boundaries of video generation.

Capabilities

Negative PromptSeed

Cost

Community model (estimated from hardware time)

Input Parameters

Name	Type	Description	Default	Constraints
`prompt`*	string	Text prompt for image generation	`—`	—
`aspect_ratio`	string	Aspect ratio of the output video.	`"16:9"`	16:99:16
`disable_safety_checker`	boolean	Disable safety checker for generated videos	`false`	—
`fast_mode`	string	Speed up generation with different levels of acceleration. Faster modes may degrade quality somewhat. The speedup is dependent on the content, so different videos may see different speedups.	`"Balanced"`	OffBalancedFast
`lora_scale`	number	Determines how strongly the main LoRA should be applied. Sane results between 0 and 1 for base inference. You may still need to experiment to find the best value for your particular lora.	`1`	min: 0, max: 4
`lora_weights`	string	Load LoRA weights. Supports HuggingFace URLs in the format huggingface.co/<owner>/<model-name>, CivitAI URLs in the format civitai.com/models/<id>[/<model-name>], or arbitrary .safetensors URLs from the Internet.	`—`	—
`negative_prompt`	string	Negative prompt to avoid certain elements	`""`	—
`sample_guide_scale`	number	Guidance scale for generation	`5`	min: 1, max: 10
`sample_shift`	integer	Flow shift parameter for video generation	`3`	min: 0, max: 10
`sample_steps`	integer	Number of inference steps	`30`	min: 1, max: 40
`seed`	integer	Random seed. Set for reproducible generation	`—`	—

promptrequiredstring

Text prompt for image generation

aspect_ratiostring

Aspect ratio of the output video.

Default: "16:9"

16:99:16

disable_safety_checkerboolean

Disable safety checker for generated videos

Default: false

fast_modestring

Speed up generation with different levels of acceleration. Faster modes may degrade quality somewhat. The speedup is dependent on the content, so different videos may see different speedups.

Default: "Balanced"

OffBalancedFast

lora_scalenumber

Determines how strongly the main LoRA should be applied. Sane results between 0 and 1 for base inference. You may still need to experiment to find the best value for your particular lora.

Default: 1min: 0, max: 4

lora_weightsstring

Load LoRA weights. Supports HuggingFace URLs in the format huggingface.co/<owner>/<model-name>, CivitAI URLs in the format civitai.com/models/<id>[/<model-name>], or arbitrary .safetensors URLs from the Internet.

negative_promptstring

Negative prompt to avoid certain elements

Default: ""

sample_guide_scalenumber

Guidance scale for generation

Default: 5min: 1, max: 10

sample_shiftinteger

Flow shift parameter for video generation

Default: 3min: 0, max: 10

sample_stepsinteger

Number of inference steps

Default: 30min: 1, max: 40

seedinteger

Random seed. Set for reproducible generation

Version: f5576aa30f5dUpdated: 7/25/202635.5K runs