text-to-image

Text To Image

Overview

Generate brand-new images from scratch based on text descriptions (no base image); reference images only supplement style/atmosphere rather than serving as the main content. Covers 2 types of scenarios:

1. General text-to-image / multi-image reference fusion: creative concept art, illustrations, style mashups, fast image generation; supports image_list multi-image reference fusion; the default channel that is divergent/diverse/fast.
2. Commercial-grade high-quality generation: promotional visuals/covers, e-commerce posters/marketing banners, product scene images, text rendering (infographic/poster text), photorealistic/movie still style; deep text understanding, accurate text rendering, high resolution;Does not support image_list。

Text-to-image with descriptions of people/characters/portraits (character covers, concept art, illustrations containing people) is handled by this tool. Does not handle: editing/modifying based on existing images, reference-image-driven creation (where the reference image is the main source of content),reference image already providedhigh-fidelity portraits/pet photos/ID photos, poster layout design, style transfer, video generation, image search, image quality enhancement/super-resolution.

API Mapping

Scenario	Backend API
General text-to-image / multi-image reference fusion	image_gummy_generate
Commercial-grade high-quality generation	image_praline_create_v2
Commercial-grade fallback	image_praline_create_2(fallback channel when praline_v2 fails, parameters compatible)
Nougat high quality (explicitly specified externally)	image_nougat_create(does not support image_list / output_format; quality defaults internally in the API to Medium and is not exposed externally)

model mapping:

• auto(default) / not passed → decided according to routing rules
• gummy → image_gummy_generate
• praline_pro → image_praline_create_v2
• nougat → image_nougat_create(triggered only when explicitly specified externally; not recognized by routing)
• When a call with a specified model fails, retries continue according to the fallback chain

Dependencies

• meitu-cli: >=2.0.6
• Credentials：CONFIG AKSK → meitu tools update; EXEC AKSK → run command (see root CONFIG.md）
• Environment variables：MEITU_OPENAPI_TOOL_TASK_MODE=command

Core Workflow

Preflight → Execute → Deliver

Preflight

1. meitu --version ≥ 2.0.6
2. meitu auth verify --json
3. Confirm registry contains text-to-image
4. Resolve output_dir:openclaw.yaml → ./output/；else → ~/.openclaw/workspace/visual/output/text-to-image/

Execute

Trigger signals / routing rules

Decision order:

1. Has image_list → force image_gummy_generate(praline does not support image_list)
2. No image_list: commercial/promotional/text rendering/photorealistic movie style → image_praline_create_v2; otherwise → image_gummy_generate(default)

Scenario	Judgment keywords	Routing
Has reference image (multi-image fusion)	refer to this image, fusion, multiple images	image_gummy_generate
Commercial-grade / promotional / e-commerce	promotional image, cover, e-commerce, marketing, Banner, commercial-grade, high quality, product scene image	image_praline_create_v2
Text rendering	with text, add title, infographic, poster text	image_praline_create_v2
Photorealistic / movie style	photorealistic, realistic feel, light and shadow texture, movie still	image_praline_create_v2
Creative / illustration / concept	draw an image, concept art, illustration, creative, style mashup	image_gummy_generate
General / no special requirements	Other	image_gummy_generate(Default)

Parameter Definitions

Common Parameters:

Parameter	Type	Required	Range	Default Value	Description
prompt	STRING	Yes	--	--	Image description. Missing → follow-up question
size	STRING	No	1K / 2K / 4K	2K	Output resolution

image_gummy_generate Additional Parameters:

Parameter	Type	Required	Default Value	Description
image_list	ARRAY[STRING]	No	--	Array of reference image URLs, for multi-image fusion/style reference; not passed for pure text-to-image

image_praline_create_v2 Additional Parameters:

Parameter	Type	Required	Range	Default Value	Description
ratio	STRING	No	1:1 / 2:3 / 3:2 / 3:4 / 4:3 / 4:5 / 5:4 / 9:16 / 16:9 / 21:9	--	Output aspect ratio; special ratios must be written at the end of prompt ending with Aspect ratio x:y

image_praline_create_2(fallback):

Parameter	Type	Required	Range	Default Value	Description
ratio	STRING	No	auto / 1:1 / 2:3 / 3:2 / 3:4 / 4:3 / 4:5 / 5:4 / 9:16 / 16:9 / 21:9	auto	Output aspect ratio

image_nougat_create(explicit model):

Parameter	Type	Required	Range	Default Value	Description
ratio	STRING	No	1:1 / 2:3 / 3:2 / 3:4 / 4:3 / 4:5 / 5:4 / 9:16 / 16:9 / 21:9	--	Only effective when model=nougat ; does not support image_list

Description:image_list and ratio are mutually exclusive;output_format is internally fixed by combo-tool to jpeg, not exposed externally.

Tool Invocation

Creative / Default:

meitu text-to-image \
  --skill_name skill_text-to-image \
  --prompt "{image_description}" \
  --size 2K \
  --json --download-dir {output_dir}

Multi-image reference fusion:

meitu text-to-image \
  --skill_name skill_text-to-image \
  --prompt "{image_description}" \
  --model gummy \
  --image_list "{ref1}" --image_list "{ref2}" \
  --json --download-dir {output_dir}

Commercial-grade high quality (with aspect ratio):

meitu text-to-image \
  --skill_name skill_text-to-image \
  --prompt "{image_description} Aspect ratio 16:9" \
  --model praline_pro \
  --ratio 16:9 --size 2K \
  --json --download-dir {output_dir}

Error Fallback

Scenario	Handling
prompt missing	Prompt "Please describe the image content you want to generate"
size invalid value	Fall back to 2K
image_list partially invalid	Silently filter out invalid items
image_list all invalid + has prompt	Degrade to pure text-to-image, go through image_gummy_generate
Has reference images but routed to praline	Force switch to image_gummy_generate, inform "commercial-grade mode does not support reference images"
image_praline_create_v2 failed	Fall back to image_praline_create_2 retry
image_praline_create_2 failed	Fall back to image_gummy_generate retry
image_gummy_generate failed	Retry once, return error if still fails
Content compliance interception	Directly return compliance prompt, no retry, no fallback

Deliver

Parse --json：

• ok: true → downloaded_files[0].saved_path as local image
• ok: false → output code + hint

Write to disk:mv {file} {output_dir}/{YYYY-MM-DD}_{descriptive-name}.jpg

Output

• Format: JPEG (internally fixed by combo-tool)
• Naming: {YYYY-MM-DD}_{descriptive-name}.jpg
• Location: project → ./output/; one-off → ~/.openclaw/workspace/visual/output/text-to-image/

Baseline Task ID

See root directory references/task-id-baseline.md in text-to-image entry.