twha Video Generation Model

Discover the twha Video Generation Model, a powerful tool for creating short, dynamic videos from text or images. Try it now and see the results!

Platform: Replicate

Text-to-VideoImage-to-VideoLoRA Video SynthesisShort Video Generation

3 runs

H100

License Check Required

🚀Function Overview

Generates short videos (1-5 seconds) from textual prompts and optional starting images, with adjustable quality and stylistic controls.

Key Features

Text-driven video generation
Image-based video initiation
LoRA strength adjustment for model and CLIP
Aspect ratio and resolution configuration
Frame count control (1-5 second videos)
Speed/quality trade-off via fast modes
Seed-based reproducibility

Use Cases

•Creating animated sequences from text descriptions
•Transforming still images into short video clips
•Generating social media content visuals
•Prototyping motion graphics from concepts
•Producing stylized animations with LoRA controls

⚙️Input Parameters

prompt

string

Text prompt for video generation

negative_prompt

string

Things you do not want to see in your video

image

string

Image to use as a starting frame for image to video generation.

aspect_ratio

string

The aspect ratio of the video. 16:9, 9:16, 1:1, etc.

frames

integer

The number of frames to generate (1 to 5 seconds)

resolution

string

The resolution of the video. 720p is not supported for 1.3b.

lora_strength_model

number

Strength of the LORA applied to the model. 0.0 is no LORA.

lora_strength_clip

number

Strength of the LORA applied to the CLIP model. 0.0 is no LORA.

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.

sample_steps

integer

Number of generation steps. Fewer steps means faster generation, at the expensive of output quality. 30 steps is sufficient for most prompts

sample_guide_scale

number

Higher guide scale makes prompt adherence better, but can reduce variation

sample_shift

number

Sample shift factor

seed

integer

Set a seed for reproducibility. Random by default.

💡Usage Examples

Example 1

Input Parameters

{
  "frames": 81,
  "prompt": "twha . wearing an unbuttoned men's dress shirt and panties is walking through a kitchen. she holds her morning coffee mug in her hands. as she walks the unbuttoned shirt sways with her movement, teasing the viewer with a glimpse of her breasts. the scene is almost like a fantasy you have in a dream of a woman you want to be with. set in a modern cozy kitchen.",
  "fast_mode": "Balanced",
  "resolution": "720p",
  "aspect_ratio": "16:9",
  "sample_shift": 8,
  "sample_steps": 30,
  "negative_prompt": "",
  "lora_strength_clip": 1,
  "sample_guide_scale": 5,
  "lora_strength_model": 1
}

Output Results

https://replicate.delivery/xezq/6WcdL4ttNTp5LVikLFkBgO7j4URkfZ4AgMhBiFkWdK6EwyYKA/R8_Wan_00001.mp4

Quick Actions

Use NowView Documentation

Technical Specifications

Hardware Type: H100
Run Count: 3
Commercial Use: Unknown/Restricted
Platform: Replicate

Related Keywords

Text-to-VideoImage-to-VideoLoRA Video SynthesisSocial Media ContentMotion Graphics PrototypingStylized AnimationsReproducible Video

Related Models

shridharathi/van-gogh-vid

Make your videos van gogh-esque

PixVerse Video Generator

Quickly make 5s or 8s videos at 540p, 720p or 1080p. It has enhanced motion, prompt coherence and handles complex actions well.

Lightricks LTX Video Generation

DiT-based 13b video generation model, creating 30fps video