image-portrait-generate

Image Portrait Generate

Overview

Generate high-precision portraits from reference images (human portraits/pets), covering single-person/pet/multi-person portraits (ID photos/profile photos/artistic photos/hairstyle adjustment/scene changes) and two-person composite photo synthesis. Supports 2K/4K resolution. Triggered only when the user explicitly requests portrait generation or two-person composite photo synthesis; not for generalized "group photo" or "team photo" chit-chat scenarios.

API Mapping

• Single-person/pet/multi-person portraits (primary):image_gummy_edit_v45
• Single-person/pet portrait fallback (last resort):image_mint_portrait_edit_2k
• Two-person composite photo synthesis:image_groupphoto2_edit
• model mapping:auto(default)/gummy_pro→v45 / mint_portrait→mint_2k

Dependencies

• meitu-cli: >=2.0.6
• credentials：CONFIG AKSK → meitu tools update; EXEC AKSK → actual execution (see root CONFIG.md）
• environment variables：MEITU_OPENAPI_TOOL_TASK_MODE=command

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

Core Workflow

Preflight → Execute → Deliver

Preflight

1. meitu --version ≥ 2.0.6
2. Already ran with CONFIG AKSK meitu tools update
3. current AKSK = EXEC,MEITU_OPENAPI_TOOL_TASK_MODE=command
4. output_dir：openclaw.yaml → ./output/ ｜else → $VISUAL/output/image-portrait-generate/；mkdir -p

Execute

trigger signals / routing rules

core dimensions:number of input images + group photo signal。

scenario	determination keywords	number of images	routing
two-person composite photo	group photo, composite photo, two people composite	exactly 2	image_groupphoto2_edit
two-person composite photo but fewer than 2 images	group photo	1	prompt "please provide two portrait photos"
two-person composite photo but more than 2 images	group photo	>2	take the first 2 images + inform
single-person / pet portrait	portrait, ID photo, profile photo, hairstyle, pet portrait	1 or >1	image_gummy_edit_v45 → image_mint_portrait_edit_2k
multi-person portrait (3+ people)	team photo, 3-person portrait	1 (multi-person image)	image_gummy_edit_v45(no fallback)
unclear intent + 2 images	--	2	default to single-person portrait

parameter definitions

parameter	type	required	range	default	description
image_list	ARRAY[STRING]	yes	1–N images	--	array of portrait/pet image URLs
prompt	STRING	yes	--	--	portrait description; automatically generated from context when missing
size	STRING	no	2K / 4K	2K	output resolution (only image_gummy_edit_v45 supported)

API mapping description:

• image_gummy_edit_v45(primary):image_list[0]→image_url(the API itself supports multiple people),prompt/size direct pass
• image_mint_portrait_edit_2k(fallback):image_list[0]→image_url，prompt direct pass
• image_groupphoto2_edit(two-person composite photo):image_list array (must be exactly 2),prompt→instruction(non-Chinese/English automatically translated to Chinese); fixed parameter lora_key=multi_duorenhezhao_2k、rsp_media_type=url

ID photo + specified background color (blue/white/red): write the background color requirement into prompt (e.g., "blue solid background ID photo").

tool invocation

meitu image-portrait-generate --image_list <img1>[,<img2>,...] --prompt "<desc>" [--size 2K|4K] --json --download-dir {output_dir} --skill_name skill_image-portrait-generate

error fallback

scenario	handling method
image_list missing	prompt "please provide a reference image", do not call API
image_list any image in is inaccessible	Return an invalid image link error, do not retry
prompt missing	Automatically generated based on context
Group photo but only 1 image	Prompt "Please provide two portrait photos"
Group photo and >2 images	Take the first 2 images and notify
Portrait but 2 images provided	Default to single-person portrait
image_gummy_edit_v45 failed (single person/pet)	Automatically downgrade to image_mint_portrait_edit_2k
image_mint_portrait_edit_2k failed	Return error
image_gummy_edit_v45 failed (multiple people/3+ people)	Do not downgrade, return error directly
image_groupphoto2_edit failed	Return error
No face/pet detected	Return error: reference image must contain a clear portrait or pet
Content compliance interception	Return compliance prompt directly, do not retry, do not downgrade

Deliver

• Use the output_dir parsed by Preflight
• From downloaded_files[0].saved_path read the downloaded file path
• mv {downloaded_files[0].saved_path} {output_dir}/{YYYY-MM-DD}_{descriptive}_image-portrait-generate.{ext}

Output

• Format：image
• Location: project → ./output/, one-time → $VISUAL/output/image-portrait-generate/

Baseline Task ID

See references/task-id-baseline.md corresponding row in.