An AI Agent Skill for generating videos using Vidu models via Atlas Cloud API. Text-to-video, image-to-video, start-end frame video, reference-to-video — all from your terminal. Anime style, 1080p, audio generation, and fine-grained motion control.
Built for the open agent skills ecosystem — works with Claude Code, Cursor, Codex, Copilot, Gemini CLI, Windsurf, OpenCode, Kiro, and 15+ AI coding agents.
- Text-to-Video — Generate video from a text prompt
- Image-to-Video — Animate a static image into a video
- Start-End Frame Video — Define start and end frames, Vidu fills the motion between
- Reference-to-Video — Use a reference image to guide style and character consistency
- Anime Style — Built-in anime mode for stylized generation
- 1080p Output — Full HD resolution support
- Audio Generation — Optional AI-generated audio and background music
- Movement Amplitude Control — Fine-tune motion intensity (auto, small, medium, large)
- Multiple Aspect Ratios — 16:9, 9:16, 1:1, and more
- Variable Duration — 5s or 8s video clips
- Affordable Pricing — Starting at just from $0.034/s with Atlas Cloud
| Model | Mode | Starting Price per Second | Resolution | Duration | Notes |
|---|---|---|---|---|---|
| Vidu Q3-Pro T2V | Text-to-Video | from $0.06/s | Up to 1080p | 5/8s | Highest quality |
| Vidu Q3-Pro I2V | Image-to-Video | from $0.06/s | Up to 1080p | 5/8s | Animate any image |
| Vidu Q3-Pro Start-End | Start-End Frame | from $0.06/s | Up to 1080p | 5/8s | Define start and end frames |
| Vidu Q3-Turbo T2V | Text-to-Video | from $0.034/s | Up to 720p | 5/8s | Fast and affordable |
| Vidu Q3-Turbo I2V | Image-to-Video | from $0.034/s | Up to 720p | 5/8s | Fast image animation |
| Vidu Q3-Turbo Start-End | Start-End Frame | from $0.034/s | Up to 720p | 5/8s | Fast start-end generation |
| Vidu Ref-to-Video Q1 | Reference-to-Video | from $0.06/s | Up to 720p | 5/8s | Character-consistent generation |
| Vidu Ref-to-Video 2.0 | Reference-to-Video | from $0.06/s | Up to 720p | 5/8s | Improved reference adherence |
Prices shown are starting prices. Higher resolution or longer duration may cost more.
⚠️ Note: Vidu Q3 models may add mosaic/blur to certain NSFW scenes due to training data limitations. Not guaranteed 100% uncensored. For reliable uncensored output, use Wan 2.2 Spicy ($0.03) or Wan 2.6 ($0.07).
Requirements: Bun
# Clone the repo
git clone https://github.com/thoughtincode/vidu-video-skill.git ~/tools/vidu-video-skill
cd ~/tools/vidu-video-skill
# Install dependencies
bun install
# Link globally (no sudo needed - uses Bun's global bin)
bun link
# Set up your API key
cp .env.example .env
# Edit .env and add your Atlas Cloud API keyGet an Atlas Cloud API key at Atlas Cloud.
Now you can use vidu-video from anywhere.
When installed as an agent skill (works with Claude Code, Cursor, Codex, Copilot, and 15+ AI coding agents), just say /init and your AI agent will clone the repo, install deps, and link the command for you. Then use it by saying "generate a video of..." and the agent handles the rest.
mkdir -p ~/.local/bin
ln -sf ~/tools/vidu-video-skill/src/cli.ts ~/.local/bin/vidu-video
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc
source ~/.zshrc# Basic text-to-video
vidu-video "a cat walking across a sunny garden"
# Custom output name
vidu-video "ocean waves crashing on rocks" -o waves
# 1080p resolution
vidu-video "futuristic city flyover" --resolution 1080p
# 8-second clip
vidu-video "time-lapse of flowers blooming" --duration 8
# Custom output directory
vidu-video "sunset timelapse" -o sunset -d ~/Videos# Text-to-Video (default)
vidu-video "your prompt"
# Image-to-Video — animate a static image
vidu-video "make this character walk forward" --mode i2v --image character.png
# Start-End Frame — define start and end, Vidu fills the motion
vidu-video "smooth transition between scenes" --mode start-end --image start.png --end-image end.png
# Reference-to-Video — character/style-consistent generation
vidu-video "the character is dancing in a garden" --mode ref --image reference.png# Default — Vidu Q3-Pro (highest quality)
vidu-video "your prompt"
# Q3-Turbo — fast and affordable
vidu-video "your prompt" --model q3-turbo
# Q3-Pro Image-to-Video
vidu-video "animate this scene" --model q3-pro-i2v --image scene.png
# Reference-to-Video 2.0
vidu-video "character walks through forest" --model ref-2.0 --image character.png| Alias | Model ID | Best For |
|---|---|---|
q3-pro, q3-pro-t2v |
vidu/q3-pro/text-to-video |
Best quality text-to-video |
q3-pro-i2v |
vidu/q3-pro/image-to-video |
Animating static images |
q3-pro-start-end |
vidu/q3-pro/start-end-to-video |
Start-end frame interpolation |
q3-turbo, q3-turbo-t2v |
vidu/q3-turbo/text-to-video |
Fast, cost-effective generation |
q3-turbo-i2v |
vidu/q3-turbo/image-to-video |
Fast image animation |
q3-turbo-start-end |
vidu/q3-turbo/start-end-to-video |
Fast start-end generation |
ref-q1 |
vidu/reference-to-video/q1 |
Character-consistent video |
ref-2.0 |
vidu/reference-to-video/2.0 |
Improved reference adherence |
Use the --style anime flag to generate anime-styled videos:
# Anime text-to-video
vidu-video "a magical girl transformation sequence" --style anime
# Anime image-to-video
vidu-video "animate this anime character running" --style anime --mode i2v --image character.pngThe --style flag supports general (default) and anime.
Control the intensity of motion in generated videos:
# Auto (model decides)
vidu-video "a person talking" --movement auto
# Minimal motion
vidu-video "a serene lake at dawn" --movement small
# Moderate motion
vidu-video "a person walking through a park" --movement medium
# Maximum motion
vidu-video "an intense car chase scene" --movement largeGenerate AI audio and background music for your videos:
# Enable audio generation
vidu-video "a thunderstorm rolling in" --audio
# Enable background music
vidu-video "a romantic sunset scene" --bgm
# Both audio and music
vidu-video "a busy city street" --audio --bgmDefine the start and end frames of your video, and Vidu fills in the motion between them:
# Provide start and end frames
vidu-video "smooth camera pan from left to right" --mode start-end --image start.png --end-image end.png
# With custom duration
vidu-video "character walks from A to B" --mode start-end --image pose-a.png --end-image pose-b.png --duration 8# 540p (fastest)
vidu-video "quick concept video" --resolution 540p
# 720p (default, balanced)
vidu-video "product demo video" --resolution 720p
# 1080p (highest quality, Q3-Pro only)
vidu-video "cinematic trailer" --resolution 1080p
# 5-second clip (default)
vidu-video "logo animation" --duration 5
# 8-second clip
vidu-video "extended scene" --duration 8# Widescreen (default)
vidu-video "cinematic landscape" --aspect 16:9
# Portrait (social media)
vidu-video "mobile app promo" --aspect 9:16
# Square
vidu-video "instagram post" --aspect 1:1| Option | Default | Description |
|---|---|---|
-o, --output |
vidu-gen-{timestamp} |
Output filename (no extension) |
--model |
q3-pro |
Model alias or full model ID |
--mode |
t2v |
Generation mode: t2v, i2v, start-end, ref |
--duration |
5 |
Video duration: 5 or 8 seconds |
--resolution |
720p |
Output resolution: 540p, 720p, 1080p |
--style |
general |
Visual style: general or anime |
--movement |
auto |
Movement amplitude: auto, small, medium, large |
--audio |
false |
Enable AI audio generation |
--bgm |
false |
Enable AI background music |
--image |
- | Input/start image path (for i2v/start-end/ref modes) |
--end-image |
- | End frame image path (for start-end mode) |
--aspect |
16:9 |
Aspect ratio: 16:9, 9:16, 1:1, etc. |
-d, --dir |
current directory | Output directory |
--api-key |
- | Atlas Cloud API key (overrides env/file) |
-h, --help |
- | Show help |
The CLI resolves the Atlas Cloud API key in priority order:
--api-keyflag on the command lineATLAS_API_KEYenvironment variable.envfile in the current working directory.envfile in the repo root (next tosrc/)~/.vidu-video/.env
# Option 1: Environment variable
export ATLAS_API_KEY=your_key_here
# Option 2: .env file in current directory
echo "ATLAS_API_KEY=your_key_here" > .env
# Option 3: Global config
mkdir -p ~/.vidu-video
echo "ATLAS_API_KEY=your_key_here" > ~/.vidu-video/.env
# Option 4: Pass directly
vidu-video "your prompt" --api-key your_key_hereThe CLI uses the Atlas Cloud API to interface with Vidu's video generation models:
- Submit Request — Sends your prompt, images, and configuration to the Atlas Cloud API
- Poll for Completion — Checks the prediction status every 5 seconds until the video is ready
- Download Result — Downloads the generated video to your specified output location
POST /api/v1/model/prediction
→ Returns request_id
GET /api/v1/model/prediction/{request_id}
→ Poll until status: "completed"
→ Download output video URL
Default: vidu/q3-pro/text-to-video
--mode i2v: vidu/q3-pro/image-to-video
--mode start-end: vidu/q3-pro/start-end-to-video
--mode ref: vidu/reference-to-video/2.0
--model q3-turbo: vidu/q3-turbo/text-to-video
--model q3-turbo-i2v: vidu/q3-turbo/image-to-video
--model ref-q1: vidu/reference-to-video/q1
Vidu's flagship model delivering the highest quality video generation. Q3-Pro supports full 1080p resolution and excels at complex prompts, detailed scenes, and photorealistic output. It handles all generation modes: text-to-video, image-to-video, and start-end frame video.
- Best for: Production-quality video, commercial content, high-detail scenes
- Strengths: 1080p output, superior motion quality, accurate prompt following
- Price: from $0.06/s
The fast and affordable variant optimized for speed and cost efficiency. Q3-Turbo generates videos up to 720p with shorter processing times, making it ideal for rapid prototyping, batch generation, and cost-sensitive production workflows.
- Best for: Prototyping, batch generation, fast iteration, budget-conscious production
- Strengths: Fast generation, low cost, reliable quality
- Price: from $0.034/s
⚠️ Note: Vidu Q3 models may add mosaic/blur to certain NSFW scenes due to training data limitations. Not guaranteed 100% uncensored. For reliable uncensored output, use Wan 2.2 Spicy ($0.03) or Wan 2.6 ($0.07).
Specialized models for character-consistent and style-consistent video generation. Provide a reference image of a character or scene, and the model maintains visual consistency throughout the generated video.
- Q1: First generation reference model with solid character consistency
- 2.0: Improved adherence to reference images with better motion quality
- Best for: Character animation, consistent branding, series content
- Strengths: Visual consistency, character preservation, style transfer
- Price: from $0.06/s
Vidu's start-end frame mode is unique among video generation models. You provide the first and last frames, and the model generates smooth, coherent motion between them. This is perfect for:
- Morphing effects — Smooth transitions between two different images
- Controlled motion — Define exact start and end poses for character animation
- Scene transitions — Create cinematic transitions between shots
- Storyboard-to-video — Animate between storyboard keyframes
Fine-tune how much motion appears in your videos with the --movement parameter:
- auto — Model decides the appropriate level of motion
- small — Subtle, minimal movement (talking heads, gentle breeze)
- medium — Moderate motion (walking, slow panning shots)
- large — Dynamic, high-energy motion (action scenes, fast camera movement)
Built-in anime style support produces high-quality anime-styled videos without needing to craft complex style prompts. The --style anime flag optimizes the model's output for anime aesthetics, including:
- Character design with anime proportions
- Anime-style shading and coloring
- Fluid animation consistent with anime conventions
- Background art in anime style
- Marketing Videos — Product demos, social media clips, ad creatives
- Anime Content — Anime-styled videos with built-in style support
- Character Animation — Reference-to-video for consistent character videos
- Scene Transitions — Start-end frame mode for smooth transitions
- Prototyping — Q3-Turbo for quick, affordable video concepts
- Social Media — Portrait mode (9:16) for TikTok, Reels, Shorts
- Audio-Visual Content — Built-in audio and BGM generation
- Storyboard Animation — Start-end frame mode to animate storyboards
- Game Development — Cutscene concepts, character motion studies
- Video Production — Visual elements for Remotion/video compositions
When installed as an agent skill (works with Claude Code, Cursor, Codex, Copilot, Gemini CLI, Windsurf, Kiro, and more), the skill triggers on phrases like:
- "generate a video"
- "create a video clip"
- "animate this image"
- "make an anime video"
- "create a transition between these frames"
- "generate a video with audio"
Your AI agent will construct the appropriate vidu-video command based on your request, handling model selection, resolution, duration, style, movement, audio, and output configuration automatically.
User: "Generate an 8-second anime video of a samurai battle"
Agent: vidu-video "epic samurai battle scene, dynamic sword fighting" --style anime --duration 8
User: "Animate this character image with minimal movement"
Agent: vidu-video "character performs subtle idle animation" --mode i2v --image character.png --movement small
User: "Create a smooth transition between these two frames"
Agent: vidu-video "smooth cinematic transition" --mode start-end --image start.png --end-image end.png
User: "Make a product demo video with background music"
Agent: vidu-video "product rotating on pedestal, clean white background" --bgm --resolution 1080p
| Feature | Vidu Q3-Pro | Wan 2.6 | Kling 2.0 | Runway Gen-3 |
|---|---|---|---|---|
| Price | from $0.06/s | from $0.07/s | $0.10+ | $0.50+ |
| 1080p Support | Yes | No | Yes | Yes |
| Anime Mode | Yes | No | No | No |
| Start-End Frame | Yes | No | No | No |
| Reference Video | Yes | No | No | No |
| Audio Generation | Yes | No | No | No |
| Movement Control | Yes | No | No | No |
| Text-to-Video | Yes | Yes | Yes | Yes |
| Image-to-Video | Yes | Yes | Yes | Yes |
Make sure your Atlas Cloud API key is set. See API Key Configuration.
Video generation can take 30-120 seconds depending on the model and resolution. The CLI will poll automatically. If it times out after 5 minutes, try again or use a lower resolution.
Check that you're using a valid model alias. Run vidu-video --help to see all available models.
Both start and end frame images are required for start-end frame mode.
Q3-Turbo supports up to 720p. Use --model q3-pro for 1080p output.
This workflow is optimized for Atlas Cloud. Move from experiment to enterprise-ready scale.
- Production-Ready: Vidu Q3-Pro at only from $0.06/s — premium quality at low cost
- Budget Option: Vidu Q3-Turbo at just from $0.034/s for high-volume generation
- Enterprise Security: SOC I & II Certified | HIPAA Compliant
- Zero Maintenance: Serverless architecture—focus on your product, not the servers
MIT