OpenClaw Selfie — Identity-Consistent Photo Generation for OpenClaw Characters

What This Skill Does

Generate photos where your OpenClaw character looks the same every time — selfies, group shots,
portraits, stylized scenes, and more — all through tuqu.ai's image generation API. The skill handles
character identity preservation, prompt enhancement, preset application, balance management, and
billing.

Installation

Install via ClawHub

CODEBLOCK0

Or with the native OpenClaw command:

CODEBLOCK1

Prerequisites

• - Python 3.8+ must be available on your system (python3 in PATH).
• A tuqu.ai service key for each OpenClaw role that needs to generate images. Each role can have

its own key — the skill passes it explicitly per request via --service-key.

Post-Install Verification

After installation, verify the skill is working:

CODEBLOCK2

If both commands return JSON, the skill is ready. Authenticated calls (generation, balance, etc.)
also require --service-key <your-role-service-key>.

Optional Environment Overrides

Only set these when you need to point at a non-default host:

• - TUQU_BASE_URL — defaults to INLINECODE4
• INLINECODE5 — defaults to INLINECODE6

How It Works

All API calls go through scripts/tuqu_request.py. The helper auto-selects the correct host and
auth mode for each endpoint, keeps credentials explicit via --service-key, and prints formatted
JSON for direct inspection.

Detailed API semantics are in TUQU_API.md. Exact request/response fields are in
references/endpoints.md and task sequences in
references/workflows.md.

Important: Credential Handling

Authenticated calls must pass --service-key <role-service-key> explicitly on every request. Do not
rely on a shared credential environment variable — different OpenClaw roles can carry different keys.

Use These Command Patterns

List or query data:

CODEBLOCK3

Send a small JSON body inline:

CODEBLOCK4

Send a larger payload from disk:

CODEBLOCK5

Override helper defaults only with a documented reason:

CODEBLOCK6

Run Supported Tasks Through the Helper

Classify the Request First

Before picking an endpoint, classify the user request into one of these buckets:

1. 1. Current-role selfie or portrait request:

自拍, 照片, 写真, 发张图, or similar wording that implies the current role should be in frame

1. 2. Character-on-camera request:

the user explicitly wants the current role or a saved character to appear in the image

1. 3. Freestyle or edit-only request:

landscape, objects, scenery, atmosphere shots, or pure image editing without the current role

If the request is ambiguous, decide whether the current role needs to appear in the final image.

Decide Whether the Current Role Must Appear

• - Treat 自拍 as current-role-on-camera by default.
• INLINECODE15 means the current role appears in the image. It does not mean a phone must be visible.
• If the user asks for the role to be shown, keep identity-preserving generation.
• If the request is for scenery, objects, mood boards, or edit-only transforms, do not force the

current role into the frame.

• - Do not ask the user for their own face photo unless they explicitly ask to put themselves in the

image.

Route by Subject Type

Use identity-preserving routing when the current role must appear:

• - Selfie / portrait / role-on-camera requests -> INLINECODE16
• Freestyle / landscape / object / edit-only requests -> INLINECODE17

Keep all supported calls on scripts/tuqu_request.py.

Run Character Prechecks Before Identity-Preserving Generation

When the current role must appear in the frame, enforce this order:

1. 1. Check whether the current role already has a Tuqu character.
2. If not, create the character first through /api/characters.
3. Check balance through /api/billing/balance.
4. Only then call /api/v2/generate-for-character.

Helper sequence:

CODEBLOCK7

Use the create-character step only when the current role does not already have a usable Tuqu
character.

Apply Default Selfie Behavior

• - For a normal selfie, default to a front-camera composition with the current role in frame.
• Do not show the phone by default.
• Show the phone only when the user explicitly asks for a mirror selfie, visible phone, or similar

framing.

• - If the user only says 自拍 or 发张图, assume the goal is a natural current-role portrait rather

than a literal handheld-phone shot.

Discover presets, models, and pricing

CODEBLOCK8

Use /api/pricing-config before accepting a user-supplied model name. Match the requested model
to a real models[].id, then use that modelId in later generation payloads.

Improve a prompt

CODEBLOCK9

Generate from prompt or reference images

CODEBLOCK10

Example payloads/generate-image.json:

CODEBLOCK11

Apply a preset

CODEBLOCK12

Manage characters

CODEBLOCK13

Generate with saved characters

CODEBLOCK14

Optionally refine the scene prompt first with /api/enhance-prompt. When the request is a selfie
or other current-role portrait, make sure the character check and balance check happen before the
generation call.

Inspect history and balance

CODEBLOCK15

Start a recharge flow

CODEBLOCK16

Operating Rules

• - Use scripts/tuqu_request.py instead of ad-hoc curl for supported endpoints.
• Classify the request before routing it: current-role selfie/portrait, role-on-camera, or

freestyle/edit-only.

• - Route identity-preserving requests to /api/v2/generate-for-character.
• Route freestyle or edit-only requests to /api/v2/generate-image.
• When the current role must appear, ensure a Tuqu character exists first, then check balance,

then generate.

• - Pass --service-key on every authenticated helper call.
• Use --body-file for large JSON payloads, especially generation and preset payloads.
• Treat /api/catalog as the source of truth for presetId, preset type, and preset variables.
• Use /api/pricing-config to resolve user-supplied model names before setting modelId.
• Treat ordinary 自拍 as front-camera composition with the current role visible and the phone not

visible unless explicitly requested.

• - Do not ask the user for their face photo unless they explicitly ask to place themselves in the

image.

• - Keep raw error payloads visible when troubleshooting; the helper already prints JSON responses.
• Read TUQU_API.md before overriding helper defaults or diagnosing auth/host issues.