Skip to content

[feat] canvas: add a character four-view preset - #406

Merged
Probiotics5333 merged 1 commit into
ThinkFlowLab:mainfrom
Probiotics5333:main
Nov 19, 2026
Merged

Probiotics5333 merged 1 commit into
ThinkFlowLab:mainfrom
Probiotics5333:main

Conversation

@Probiotics5333

@Probiotics5333 Probiotics5333 commented Nov 19, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

Add a four-view image preset for character reference sheets: one landscape sheet with front, right-side and rear full-body views plus a frontal head-and-shoulders close-up. An optional "Face emphasis" toggle, shown only for this preset, asks the model to frame the front and side views from the chin down while keeping the full rear view and the close-up.

  • Backend: new four-view preset and FOUR_VIEW_FACE_EMPHASIS prompt; image-presets exposes an optional boolean faceEmphasis.
  • Canvas: the checkbox appears only for the four-view preset, and other presets drop the parameter before submission.
  • Both labels are translated, and both READMEs describe the preset.

Linked issue

  • Fixes #N included

Fixes #405

Labels

  • Labels applied: type:feat, area:canvas, modality:image

Test plan

  • Local:
    • pytest tests/test_tool_tasks.py tests/test_tool_task_history.py — 70 passed.
    • node --test tests/toolCatalog.test.ts — 11 passed.
    • npx tsc --noEmit — clean.
    • python scripts/generate_contracts.py --check — verified.
    • npm run check:i18n — 0 problems.
    • npm test — 929 of 930 passed; the remaining browser-host case fails in this environment while creating a temporary virtualenv (WinError 5) and does not touch this change.
  • Left to CI: the static, Python, frontend and browser lanes in checks.

Evidence that it works

  • Media attached

The Four-view preset and the Face emphasis toggle in the node task panel, after the dock was widened so its note fits:

Four-view preset panel

A real sheet generated by seedream-5.0-pro from that source photo with Face emphasis enabled (3456×1152): the front and side views are framed from the chin down, with the full rear view and a frontal close-up.

Real four-view output

Provider verification

  • Real-provider calls are described honestly, or none were made

One real seedream-5.0-pro generation produced the four-view sheet above, with Face emphasis enabled; no other provider call was made. The standard (non-emphasis) four-view variant, retry and adoption paths were exercised only through the mock image transport in tests, so the default sheet still needs a real run.

Bilingual UI

  • Both locales are updated

人物四视图 / Character four-view and 脸部强调 / Face emphasis are translated, and README.md and README.zh-CN.md both describe the preset.

@github-actions github-actions Bot added area:canvas Canvas, nodes and direct manipulation modality:image Image generation or editing type:feat New capability or behaviour labels Nov 19, 2026
@github-actions
github-actions Bot requested a review from allendyu November 19, 2026 09:56
@Probiotics5333
Probiotics5333 force-pushed the main branch 3 times, most recently from 2a717b8 to 678f7f7 Compare November 19, 2026 11:35
## summary

- Add a `four-view` image preset: one landscape sheet with front, right
  side and rear full-body views plus a frontal close-up.
- Add an optional `faceEmphasis` field for `image-presets`. When the preset
  is `four-view` it swaps in a prompt that frames the front and side views
  from the chin down while keeping the full rear view and close-up.
- Show the checkbox only for that preset and drop the parameter for other
  presets; translate the new labels to English.
- Widen the image-preset dock a little and let the panel note wrap, so it
  stays readable beside the new checkbox without a wide empty gap.
- Cover the preset list and output size, the prompt swap, the catalog field
  and the preset-scoped parameter in tests, and keep the existing preset
  names in both READMEs while documenting the new one.

## validation

- `pytest tests/test_tool_tasks.py tests/test_tool_task_history.py`: 70 passed.
- `node --test tests/toolCatalog.test.ts`: 11 passed.
- `python scripts/generate_contracts.py --check`: verified.
- `npm run check:i18n`: 0 problems.
- `npx tsc --noEmit`: clean.
- `npm test`: 929 of 930 passed; the remaining browser-host case fails in
  this environment while creating a temporary virtualenv (WinError 5) and
  does not touch this change.

Fixes: ThinkFlowLab#405
@Probiotics5333

Probiotics5333 commented Nov 19, 2026 •

Copy link
Copy Markdown
Collaborator Author

PASS c841c34 Fixes #405

  • no actionable findings
  • The attached real sheet was generated with Face emphasis enabled; the standard four-view variant is covered by the prompt tests only.

Ran: Release checks green on c841c341fd1000357b926e889671544b0fb79867 (static, python, frontend, browser isolated/voice/assistant, plus CodeQL and route); pytest tests/test_tool_tasks.py tests/test_tool_task_history.py — 70 passed; node --test tests/toolCatalog.test.ts — 11 passed; npx tsc --noEmit clean; python scripts/generate_contracts.py --check verified.
Not run: real-image generation of the standard four-view variant (without Face emphasis); manual browser pass of the preset beyond the attached screenshot.

Note: this is an author self-review of the current tip, posted to record that the diff was reviewed.
Coordinator: ping after this Reviewer PASS and green CI on this SHA. This skill does not merge.

@Probiotics5333
Probiotics5333 merged commit 42a674e into ThinkFlowLab:main Nov 19, 2026
13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area:canvas Canvas, nodes and direct manipulation modality:image Image generation or editing type:feat New capability or behaviour

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[feat] canvas: 新增人物四视图预设与脸部强调

1 participant