| name | jimeng-video-generation |
| description | Generate AI videos and images using the Jimeng (Dreamina) CLI. Use this skill when the user asks to create a video, animate an image, generate an image, or produce AI-generated visual media. Triggers on phrases like "generate a video of", "create an image of", "animate this image", "make a video", "text to video", "image to video". |
Jimeng Video Generation Skill
Generate AI images and videos using the Jimeng CLI (dreamina command), powered by Jimeng 4.0+ (text-to-image) and Seedance 2.0 (video generation).
Prerequisites
- Jimeng CLI installed (run
setup.sh or install manually):
curl -s https://jimeng.jianying.com/cli | bash
- Logged in to your Jimeng account:
dreamina login
- Sufficient credits (
dreamina user_credit to check balance).
Workflow
Step 1 — Understand the request
Identify what the user wants:
- Text → Image: a still image from a description
- Text → Video: a video clip from a description
- Image → Video: animate an existing image
- Image → Image: transform/restyle an image
Step 2 — Check CLI availability
which dreamina || echo "NOT_INSTALLED"
dreamina user_credit
If not installed, run bash setup.sh first.
Step 3 — Generate
Text to Image
dreamina text2image \
--prompt="<detailed prompt>" \
--ratio=<ratio> \
--resolution_type=<resolution> \
--poll=60
--ratio: 1:1 | 16:9 | 9:16 | 4:3 | 3:4--resolution_type: 1k | 2k | 4k (default: 2k)
Text to Video
dreamina text2video \
--prompt="<detailed prompt>" \
--duration=<seconds> \
--ratio=<ratio> \
--video_resolution=<res> \
--poll=120
--duration: 5 | 10 (seconds)--ratio: 16:9 | 9:16 | 1:1--video_resolution: 480P | 720P | 1080P
Image to Video
dreamina image2video \
--image <path_or_url> \
--prompt="<motion description>" \
--duration=5 \
--poll=120
Image to Image
dreamina image2image \
--images <path_or_url> \
--prompt="<transformation description>" \
--poll=60
Multi-frame to Video
dreamina multiframe2video \
--first_frame <path> \
--last_frame <path> \
--prompt="<scene description>" \
--duration=5 \
--poll=120
Step 4 — Handle async results
If a command returns a submit_id instead of a direct result (poll timed out):
dreamina query_result --submit_id=<id>
Retry every 15–30 seconds until status is done.
Step 5 — Present results
- Show the output file path or URL to the user.
- Confirm the file was saved locally if applicable.
- Offer to refine with a follow-up generation.
Prompting Guidelines
For best results, prompts should be descriptive and specific:
Images: "<subject>, <style>, <lighting>, <composition>, <mood>"
"A lone astronaut on a red rocky planet, cinematic wide shot, golden hour lighting, dust particles in atmosphere, photorealistic"
Videos: Lead with camera movement, then subject action, then environment:
"Camera slowly pushes in, a red fox leaps through a snowy forest, slow motion, cinematic, 4K"
Image-to-video: Describe the motion/animation, not the static scene:
"Gentle camera pan left, leaves rustle softly in the breeze"
Task Management
Use dreamina list_task to view recent generation history and retrieve past results.
Troubleshooting
| Error | Fix |
|---|
not logged in | Run dreamina login or dreamina relogin |
insufficient credits | Check dreamina user_credit; top up account |
AigcComplianceConfirmationRequired | Visit jimeng.jianying.com web app and confirm content policy |
command not found: dreamina | Run bash setup.sh to install |
| Poll timeout (task still queued) | Run dreamina query_result --submit_id=<id> |
| Chrome not found (login) | Follow on-screen manual credential import instructions |
References
