Prepare a Black-and-White Portrait Edit: Source Photo, Glasses and Review
Prepare an authorized source photo, a change-and-preserve brief and one studio composition. Understand external processing, saved tasks and the visual review needed before delivering a portrait.
Fastest way
Ask your AI to do it
Paste this into the AI agent you already use:
Use Monochrome Studio Portrait to prepare an edit brief for a photo I am permitted to edit. Ask which image is the target, intended use and crop. Separate allowed changes from visible face, expression, hair and glasses details to preserve. Explain external photo processing and separate service cost before submission. Use local help for setup only; do not upload, generate or call paid APIs now. Review any actual result beside the source before delivery and keep best-effort limitations explicit.
Is this a local portrait filter?
Monochrome Studio Portrait v1.0.0 is an agent workflow for editing an authorized person photo into a quiet black-and-white half portrait. Its bundled runner uses an external image-editing service; it is not a local filter, identity verifier or proof that a photographed event occurred. Face and glasses preservation is best effort and must be reviewed against the source. Free package access does not include image-service credits.
Choose the edit target and record only visible details
Use a clear photo you are permitted to edit and identify it as the edit target. Additional angles should show the same subject and have a clear supporting role. For a minor, obtain the parent or guardian authorization required by the package. Keep the source unchanged and choose a new output filename. Do not infer private traits from appearance.
Note the visible pose, gaze, hairline, expression and skin texture. For glasses, record frame silhouette, bridge, rim thickness, fit and reflections. If only the face is visible, disclose that a half-body composition may invent unseen clothing or anatomy; prefer a modest crop and pose change.
Separate requested changes from preserved properties
Read references/portrait-contract.md and prompt-blueprints.md. The default is landscape 16:9, a chest- or waist-up crop and approximately 35–50 percent quiet negative space. Let the source pose guide the subject side. These are composition choices, not evidence of identity accuracy or a requirement for every avatar crop.
| Brief section | Write down | Review later |
|---|---|---|
| Change | Backdrop, wardrobe direction, crop and finish | Only the requested changes were made |
| Preserve | Visible face, expression, hair and glasses details | Compare source and result side by side |
| Composition | Aspect ratio, subject side and quiet space | No crop through hair, shoulders or glasses |
| Lighting and finish | One soft directional light and restrained grain | Natural texture, readable eyes and plausible reflections |
| Constraints | Unchanged approved features and no unwanted text | No invented branding or unsupported likeness guarantee |
Check the runner without uploading a photo
After installing the complete package, this command prints local usage and exits without creating a task. The v1.0.0 runner does not provide a dry-run mode or a separate --confirm-paid switch. Do not run a new edit command with a configured key until source transmission and the actual service cost are approved. A missing key prevents a new submission but is not a general sandbox or a result-quality test.
Confirm installation and invocation in your agent
node scripts/edit-portrait.mjs --helpUnderstand external processing and task recovery
For the bundled service path, FAL and the underlying model process the source photo. The runner requests private output access and limited retention, but that is not end-to-end encryption or a promise of zero temporary provider storage. Read references/runtime.md before authorizing the upload and paid edit. Keep credentials in private configuration, outside prompts, images and shared notes.
Keep the same task file and original inputs when resuming a known request. A claimed task with no request ID means the submission outcome is uncertain; do not delete it and submit again automatically. A rejected or unsatisfactory result also does not authorize another paid edit. The package protects existing output paths from overwrite; keep the original source separately as well.
Compare the actual portrait at full and thumbnail sizes
- Compare the face and glasses directly with the source. Inspect proportions, expression, hair, frame bridge, temples, lens edges and reflections. A script exit code or file hash cannot establish faithful identity preservation.
- Check eye catchlights, skin and fabric texture, shoulders and any visible hands. Look for plastic smoothing, invented accessories or an implausible head angle. Review contrast, crop, quiet space, grain and lack of unwanted words or branding at the intended use size.
- If one feature drifts, request one targeted correction while freezing the approved crop, pose, lighting and other details. Obtain new cost authorization for an external correction and review the new result again.
Deliver the result and state what remains unverified
Deliver the actual reviewed WebP, exact prompt, task record when used and a short QA note naming what was compared and any remaining differences. Keep source photos and private task files out of public repositories. Do not present an ungenerated draft as a finished portrait, use the output as an identity document or claim biometric equivalence. This guide’s offline runner checks do not certify a generated portrait.