video-content-replace

Video Content Replace

Overview

YesExisting videoPerform content element replacement (people/scenes/clothing/garments/outfits), preserving the original camera movement and actions, covering 2 types of scenarios:

1. Text instruction replacement: Text instructions drive video content modification, no additional reference image needed.
2. Reference-image driven replacement: The user provides a reference image + video, and the corresponding elements in the video are replaced according to the reference image.

Not handled: video generation, watermark/subtitle removal, overall style transformation, audio modification, video splicing.

API Mapping

Scenario	API name	model
Text instruction replacement	video_molasses_edit	molasses
Reference-image driven replacement	video_molasses_vidimgs2vid	molasses

model Optional:auto(default, automatically routed based on whether a reference image exists) / molasses(explicitly specified).

Dependencies

• meitu-cli: >=2.0.6
• Credentials：CONFIG AKSK → meitu tools update; EXEC AKSK → run command
• Environment variables：MEITU_OPENAPI_TOOL_TASK_MODE=command

Path alias: $VISUAL = {OPENCLAW_HOME}/workspace/visual/

Core Workflow

Preflight → Execute → Deliver

Preflight

1. meitu --version and meitu auth verify --json
2. Detect MEITU_OPENAPI_TOOL_TASK_MODE=command
3. output_dir: project → ./output/; one-time → $VISUAL/output/video-content-replace/；mkdir -p

Execute

Trigger signals and routing

Core judgment dimension:whether a reference image is provided。

Scenario	Judgment	Routing
Has reference image + replace element	Has image_url	video_molasses_vidimgs2vid
No reference image + text instruction	No image_url	video_molasses_edit(default)

Routing principle:image_url When it exists, prioritize the reference-image mode, even if the user says "modify using a text description", still use vidimgs2vid。

Parameter definitions

Parameter	Type	Required	Range	Default	Description
video_url	STRING	Yes	See constraints below	—	Original video address
prompt	STRING	Yes	—	—	Modification description
image_url	STRING	No	—	—	Reference image URL; when present, route to vidimgs2vid
model	STRING	No	auto/molasses	auto	Model selection

The remaining parameters (duration/ratio/character_orientation etc.) are managed internally by combo-tool and are not exposed externally.

video_url Constraints (dedicated to reference-image driven mode)：

• Format:.mp4 / .mov; size: ≤ 100MB
• Side length: 340px ~ 3850px; duration: 3 ~ 30 seconds
• Image: clearly show the upper body/full body, 1 person action, one continuous shot, stable movement

video_molasses_edit(Text instruction mode) The restrictions on video_url are relatively loose, mainly format and size.

Tool invocation

# 文本指令
meitu video-content-replace --video_url {url} --prompt "{修改描述}" --json --download-dir {output_dir} --skill_name skill_video-content-replace

# 参考图驱动
meitu video-content-replace --video_url {url} --prompt "{修改描述}" --image_url {ref} --json --download-dir {output_dir} --skill_name skill_video-content-replace

Error fallback

Scenario	Handling
video_url Missing	Prompt "Please provide a video link", do not invoke
video_url Inaccessible	Return the error directly, do not retry
prompt Missing	Prompt "Please describe the content you want to modify"
image_url Inaccessible	Fall back to video_molasses_edit(prompt only)
video_url not mp4/mov	Prompt for format requirements, do not call
video_url >100MB or edge length out of bounds	Prompt for the specific exceeded item
video_url duration <3s / >30s (vidimgs2vid)	Prompt for duration requirements
Return "Video duration can not longer than"	Identify as reference video duration exceeding the limit (not generation duration), prompt the user to trim the video rather than retry
API call failed / timed out	Retry once; if still failing, return an error
Content compliance interception	Directly return a compliance prompt, do not retry

Deliver

mv {file} {output_dir}/{date}_{name}_content-replace.mp4

Output

• Format：MP4
• Naming：{YYYY-MM-DD}_{descriptive-name}_content-replace.mp4
• Location: Project → ./output/; One-time → $VISUAL/output/video-content-replace/

Baseline Task ID

t_mt1a3i5n7b82954cc9-d2cb-44cd-a786-cb9fd37244f9