@koda-sl/baker-cli 0.432.0-dev.795979397 → 0.439.0-dev.33cf4ca99

package/README.md

@@ -4777,7 +4777,7 @@ array as a literal list: `"reference": ["$ref:subject.sheet", "$ref:type.image"]
 |---|---|---|
 | `images` | image[] | `image/png \| image/jpeg \| image/webp` (provider-determined) |
 
-**Cost** — 5 credits estimate per image asked for (`n`, default 1), 20 on Gemini 3 Pro Image (as `image_aspect_adapt` quotes it) and 20 at quality `max` (a static ad's re-laid-out sizes); the backend charges the call's actual OpenRouter usage.
+**Cost** — 5 credits estimate per image asked for (`n`, default 1), 20 on Gemini 3 Pro Image (as `image_aspect_adapt` quotes it) and 20 at quality `max`; the backend charges the call's actual OpenRouter usage.
 
 **Several candidates from one call** — `n` (integer, default 1) asks one call for several images, returned together in `images`. Each image is billed as if it were its own call (`n: 3` costs three images); what it saves is requests and waiting. Only models that declare it take it (today `openai/gpt-image-2.5-sunburst`, up to 10); any other model refuses it before anything is billed. Wire the whole array into `image_select` to keep the best.
 
@@ -5992,14 +5992,14 @@ The two scaffold passes are billed (the full `video_deconstruct` is the heavy on
 
 Turn a source/inspiration image into a **runnable, self-validated static-ad canvas** — the static counterpart of `scaffold-video`. It runs **billed Gemini passes** up front:
 
-1. **`image_describe`** (`~google/gemini-pro-latest`) — reverse-engineers the image (layout by what fills each region, copy verbatim, typefaces, capture in words, the `mechanism` and the `audience` it is built for) and writes it next to the canvas as **`prompt.json`**, an **ad spec** (`"schema": "baker-static-ad/3"`; `/2` still reads) you rewrite into the ad you want: `viewer` (who sees it and what they already know), `message` (the one thing it tells them, never printed), `copy` (each line with `lettering`: `in_scene`, written on something in the picture (`on`, `look`), or `designed`, set in the brand's face), `scene` (`description`, `look`: photo | designed | illustration, `capture` in words, `mood`, `layout`), `brand`, `colors` (each tied to an object, from the brand palette), `references` (one picture each, labelled by role: `PERSON_1`, `ANIMAL_1`, `PRODUCT_1`, `BADGE_1`, `PROP_1`), `logo` (and `logo_position`), `third_party_brands` (other brands the offer is about, a car brand on a car loan: `name`, `what`, and `logo` naming the `BADGE_n` reference of its logo; their logos may appear), an optional `direction`, the `mechanism` to rebuild for our brand, and in `/3`: `viewer.market` (`country`, `language`, `currency`), `brand_presence` (`standard` | `native`: a native disguise carries the logo as its sender and no big button), `recognition` (the kind of product a stranger must take it for at a glance, never the format), `native` (the ad is a screen people know, drawn exactly from HTML at every size by the free `static_native` node instead of painted: `x_post` (a post or a thread, with `reply_to`), `whatsapp`, `imessage`, `trustpilot`, `google_reviews`, `notes`, `slack`; on a 9:16 the app is drawn full screen as a phone shows it: status bar and app header pinned at the top, the reply bar and tab bar (or a keyboard under a chat or note) pinned at the bottom, the posts or messages as large as the clear band holds, and the frame's headline, caption and logo a card in the brand's colours under the app's header; a wall of reviews (`trustpilot`, `google_reviews`) takes up to 16 real reviews, the first two (`FEATURED_REVIEWS`) the ad's copy set whole in the safe area and the rest filling the whole frame behind them as large, sharp cards, turned and cut by its edges, because a wall wins on how many there are; a review longer than 160 characters is shown cut at a word with "…", and that cut text is the copy the review reads back; a chat (`whatsapp`, `imessage`) on a 9:16 never shrinks below about 62% of the phone scale: its earliest messages scroll up under the header, as the app does, the node's optional `scrolled` output lists the copy lines scrolled off that size, and the run review expects those lines nowhere on it; a logo file whose drawing runs off its own edge is set aside for a whole one, and a square mark where a wordmark goes is drawn larger; the run review never flags the app's own interface words (status bar, keyboard) as under Reels; the image model renders only the screen's `photos`, each a plain phone photo of ordinary people of the market; adding the block switches an existing ad on its next run, `--fix` does not apply, an invented brand handle or a rating the Knowledge does not hold (either decimal mark) is refused as `native_screen_invalid`, and a photo of ours goes in as its `image`: a native photo takes no `avatar`, since nothing would cast it), a role on each third-party brand (`partner` | `proof` | `category_signal`), a reference's `image` (wired as a picture of ours) and a person's `avatar` (the company avatar cast), and `concept` (`why_it_wins` in five layers, `load_bearing` elements with `have`/`adapt`/`missing` and `ours`, `verdict` `use`|`adapt`|`reject`, `transfer_note`; seeded from `baker winning-ads fit`). Every render keeps the load-bearing elements; `validate`/`run` refuse a `reject` or a `missing` element with nothing named (`concept_does_not_fit`) and a rating, or a promised price, share, multiple or time, the Knowledge does not hold with that unit (`proof_without_source`); a line citing a review platform gets its real badge wired. **You never write a model prompt**: every render prompt is compiled from this file for the model it runs on (short labelled sections for gpt-image, sentences with reasons for Gemini; each line quoted once; pictures numbered with one role each; colours from the brand; realism as capture cues; the safe area from its one owner; constraints last), on every `validate` and `run`, and shown on its node. A `prompt.json` written before the ad spec (the describe blueprint) is read through an adapter and keeps working. The source's own read (`source`, and the library's `reference`) stays as bookkeeping and is never sent to a model.
+1. **`image_describe`** (`~google/gemini-pro-latest`) — reverse-engineers the image (layout by what fills each region, copy verbatim, typefaces, capture in words, the `mechanism` and the `audience` it is built for) and writes it next to the canvas as **`prompt.json`**, an **ad spec** (`"schema": "baker-static-ad/3"`; `/2` still reads) you rewrite into the ad you want: `viewer` (who sees it and what they already know), `message` (the one thing it tells them, never printed), `copy` (each line with `lettering`: `in_scene`, written on something in the picture (`on`, `look`), or `designed`, set in the brand's face), `scene` (`description`, `look`: photo | designed | illustration, `capture` in words, `mood`, `layout`), `brand`, `colors` (each tied to an object, from the brand palette), `references` (one picture each, labelled by role: `PERSON_1`, `ANIMAL_1`, `PRODUCT_1`, `BADGE_1`, `PROP_1`), `logo` (and `logo_position`), `third_party_brands` (other brands the offer is about, a car brand on a car loan: `name`, `what`, and `logo` naming the `BADGE_n` reference of its logo; their logos may appear), an optional `direction`, the `mechanism` to rebuild for our brand, and in `/3`: `viewer.market` (`country`, `language`, `currency`), `brand_presence` (`standard` | `native`: a native disguise carries the logo as its sender and no big button), `recognition` (the kind of product a stranger must take it for at a glance, never the format), `native` (the ad is a screen people know, drawn exactly from HTML at every size by the free `static_native` node instead of painted: `x_post` (a post or a thread, with `reply_to`), `whatsapp`, `imessage`, `trustpilot`, `google_reviews`, `notes`, `slack`; on a 9:16 the app is drawn full screen as a phone shows it: status bar and app header pinned at the top, the reply bar and tab bar (or a keyboard under a chat or note) pinned at the bottom, the posts or messages as large as the clear band holds, and the frame's headline, caption and logo a card in the brand's colours under the app's header; a wall of reviews (`trustpilot`, `google_reviews`) takes up to 16 real reviews, the first two (`FEATURED_REVIEWS`) the ad's copy set whole in the safe area and the rest filling the whole frame behind them as large, sharp cards, turned and cut by its edges, because a wall wins on how many there are; a review longer than 160 characters is shown cut at a word with "…", and that cut text is the copy the review reads back; a chat (`whatsapp`, `imessage`) on a 9:16 never shrinks below about 62% of the phone scale: its earliest messages scroll up under the header, as the app does, the node's optional `scrolled` output lists the copy lines scrolled off that size, and the run review expects those lines nowhere on it; a logo file whose drawing runs off its own edge is set aside for a whole one, and a square mark where a wordmark goes is drawn larger; the run review never flags the app's own interface words (status bar, keyboard) as under Reels; the image model renders only the screen's `photos`, each a plain phone photo of ordinary people of the market; adding the block switches an existing ad on its next run, `--fix` does not apply, an invented brand handle or a rating the Knowledge does not hold (either decimal mark) is refused as `native_screen_invalid`, and a photo of ours goes in as its `image`: a native photo takes no `avatar`, since nothing would cast it), a role on each third-party brand (`partner` | `proof` | `category_signal`), a reference's `image` (wired as a picture of ours) and a person's `avatar` (the company avatar cast), and `concept` (`why_it_wins` in five layers, `load_bearing` elements with `have`/`adapt`/`missing` and `ours`, `verdict` `use`|`adapt`|`reject`, `transfer_note`; seeded from `baker winning-ads fit`). Every render keeps the load-bearing elements; `validate`/`run` refuse a `reject` or a `missing` element with nothing named (`concept_does_not_fit`) and a rating, or a promised price, share, multiple or time, the Knowledge does not hold with that unit, a number spelled in words included, or a superlative it does not state in the same words (`proof_without_source`); a line citing a review platform gets its real badge wired. **You never write a model prompt**: every render prompt is compiled from this file for the model it runs on (short labelled sections for gpt-image, sentences with reasons for Gemini; each line quoted once; pictures numbered with one role each; colours from the brand; realism as capture cues; the safe area from its one owner; constraints last), on every `validate` and `run`, and shown on its node. A `prompt.json` written before the ad spec (the describe blueprint) is read through an adapter and keeps working. The source's own read (`source`, and the library's `reference`) stays as bookkeeping and is never sent to a model.
 2. **element selection** (`~google/gemini-flash-latest`) — picks the **main, identity-critical** elements (the brand logo, a showcased product, a trust badge) **plus any foreground/hero person or animal** — the emotional focal point — even a generic one, because a free-generated face/muzzle reads as AI and grows artifacts; the emotional hero always gets a real-reference slot. When the advertiser's logo appears in **more than one lockup** (a square/icon **mark** and a horizontal **wordmark**), each is emitted as its **own** element (e.g. `LOGO_MARK`, `LOGO_WORDMARK`). The describe pass also records the ad's **typography** (each typeface's classification, a best-guess family, and its weight/case), which seeds the spec's `brand.typeface`. Background extras are dropped. Each element becomes a `references` entry in the spec, named by its role and naming the canvas slot it fills, so the file says which slot grounds which subject.
 
 It then scaffolds a canvas on which **the image model designs the finished ad**, holding only what `prompt.json` holds; no element is mandatory:
 
-- Up to **three variants** (`--takes`, default 3) render on the model the router picks for the concept (today `openai/gpt-image-2.5-sunburst` at quality `max`; `--gen-model` overrides), each a finished ad: the picture and every word of the copy. Each variant is **its own request** (tagged `static_variant_call`), run in parallel on its own direction (the scene as written; another picture of the same message, by the scene's look; another line leading, named by number so no word changes), and a `collect` tagged `static_variants` gathers them, on every model: one request with `n` costs the same (`n` is billed per picture) but gave fewer distinct options, so the graph never uses it. The brief quotes every line of `copy` verbatim and asks for each once; it is compiled again from `prompt.json` on every `validate` and `run`, so an edit re-keys exactly the renders it touches. The brand's **typeface** is shown to the model as a `font_specimen` set on the `designed` lines: a file in `src/brand/fonts/` (via `src/brand/tokens.json` when present), else the heading family `tokens.json` or `BRAND.md` names, fetched as a web font file; an `in_scene` line keeps its own lettering only on the real packaging, signage or screen its `on` names (handwriting on post-its, whiteboards or chalkboards, and a line naming no surface, are set as designed type). Each line's `role` sets its size: a price or stat (else the headline) leads, the `cta` is a large solid button, fine print is smallest. **The logo is optional**: the real file in `src/brand/logos/` goes to the model as a reference to reproduce once, at least a fifth of the frame wide, where it contrasts or on a small brand-colour plate, only when `prompt.json` has `"logo": true` (an older blueprint: an `advertiser_brand` entry in `brands_logos`) and the file exists; no logo is never an error, never a slot to fill and never a stand-in. Setting it false drops the logo on the next run; adding one to a canvas scaffolded without the file needs the file in `src/brand/logos/` and a re-scaffold. `prompt.json` also carries a **`brand`** block (ground, ink, accent, CTA and palette, from `src/brand/tokens.json` roles or `BRAND.md`); a colour outside it never reaches the model. The brief describes a complete ad at the master's shape and asks only for its edge margin (see below). `baker canvas set-prompt` refuses these nodes and names `prompt.json`.
+- Up to **three variants** (`--takes`, default 3) render on the model the router picks for the concept (today `openai/gpt-image-2.5-sunburst` at quality `xhigh`, about 45 s a picture against 80 s at `max`; `--gen-model` overrides), each a finished ad: the picture and every word of the copy. Each variant is **its own request** (tagged `static_variant_call`), run in parallel on its own direction (the scene as written; another picture of the same message, by the scene's look; another line leading, named by number so no word changes), and a `collect` tagged `static_variants` gathers them, on every model: one request with `n` costs the same (`n` is billed per picture) but gave fewer distinct options, so the graph never uses it. The brief quotes every line of `copy` verbatim and asks for each once; it is compiled again from `prompt.json` on every `validate` and `run`, so an edit re-keys exactly the renders it touches. The brand's **typeface** is shown to the model as a `font_specimen` set on the `designed` lines: a file in `src/brand/fonts/` (via `src/brand/tokens.json` when present), else the heading family `tokens.json` or `BRAND.md` names, fetched as a web font file; an `in_scene` line keeps its own lettering only on the real packaging, signage or screen its `on` names (handwriting on post-its, whiteboards or chalkboards, and a line naming no surface, are set as designed type). Each line's `role` sets its size: a price or stat (else the headline) leads, the `cta` is a large solid button, fine print is smallest. **The logo is optional**: the real file in `src/brand/logos/` goes to the model as a reference to reproduce once, at least a fifth of the frame wide, where it contrasts or on a small brand-colour plate, only when `prompt.json` has `"logo": true` (an older blueprint: an `advertiser_brand` entry in `brands_logos`) and the file exists; no logo is never an error, never a slot to fill and never a stand-in. Setting it false drops the logo on the next run; adding one to a canvas scaffolded without the file needs the file in `src/brand/logos/` and a re-scaffold. `prompt.json` also carries a **`brand`** block (ground, ink, accent, CTA and palette, from `src/brand/tokens.json` roles or `BRAND.md`); a colour outside it never reaches the model. The brief describes a complete ad at the master's shape and asks only for its edge margin (see below). `baker canvas set-prompt` refuses these nodes and names `prompt.json`.
 - A **`static_prefilter`** node (tagged `variant_prefilter`) drops variants that fail the deterministic checks before the judge sees them, never all of them; its **judge** (`judge_model`, `judge_prompt`) compares the rest pairwise against the compiled brief when the checks tie (one variant passes through unjudged). A canvas scaffolded with a separate `image_select` judge (tagged `pick`) has it folded into the prefilter on its next run.
-- **Every placement format is made from the chosen picture, and no size is a crop or a paste of another.** The pick goes to exact pixels as the master (an `imagemagick` node tagged `static_master`) at the lead size: the 4:5 when a feed size is delivered, else the square, else the first portrait size, rendered at the model's nearest ratio (3:4 on GPT Image) and filled to Meta's pixels. It is delivered as it is. Every other size (9:16, 1:1, 16:9, …) is **re-laid out** from the master by one request (`image_generate` node `adapt_<size>`, tagged `static_adapt_format`, on the router's `relayout` model, GPT Image 2.5 at quality `max`): the master, the variants' own product and people pictures and the logo file as references and a brief compiled from `prompt.json` for that frame on every run (`static-adapt.ts`): the exact lines and their hierarchy, the same people (same face, same clothes), product and label, logo and none the chosen ad does not show, the words and the button re-scaled for the frame, the frame stated as zones inside its safe area, never the master shrunk onto a background, set as a square or cut. A `size_<size>` node (`static_size`) brings it to Meta's exact pixels (1:1 at 1080×1080; a landscape size the same way). Every 9:16 Reels or Stories covers then passes a free `static_safe_fit` node (`safe_<size>`), which shrinks and shifts the picture just enough (to 85% at most) to bring a word or the logo a few points past the band back inside it; a layout further out passes unchanged and the run lays it out again. The 9:16 is never extended instead (the 4:5 set whole in the clear band, its scene continued to the frame) nor set over its own blurred copy: rated in October 2026, the first showed the ad small inside a continued scene ("text super tiny") and the second "looks bad". The master shape is one decision in the code (`DEFAULT_MASTER_SHAPE`: `relayout`; the `9x16_band` and `4x5_adapt` shapes, kept for the bench, make the sizes by cutting or reframing the master, and the person rating them rejected every such way; a canvas from the retired band master, which extended a square master into the 9:16, is upgraded to the re-layout graph on its next run). Placing the logo file on every size instead of having the model draw it (`logo_composite` nodes) is a scaffold option, off until it is proven on the bench.
+- **Every placement format is made from the chosen picture, and no size is a crop or a paste of another.** The pick goes to exact pixels as the master (an `imagemagick` node tagged `static_master`) at the lead size: the 4:5 when a feed size is delivered, else the square, else the first portrait size, rendered at the model's nearest ratio (3:4 on GPT Image) and filled to Meta's pixels. It is delivered as it is. Every other size (9:16, 1:1, 16:9, …) is **re-laid out** from the master by one request (`image_generate` node `adapt_<size>`, tagged `static_adapt_format`, on the router's `relayout` model, GPT Image 2.5 at quality `xhigh`): the master, the variants' own product and people pictures and the logo file as references and a brief compiled from `prompt.json` for that frame on every run (`static-adapt.ts`): the exact lines and their hierarchy, the same people (same face, same clothes), product and label, logo and none the chosen ad does not show, the words and the button re-scaled for the frame, the frame stated as zones inside its safe area, never the master shrunk onto a background, set as a square or cut. A `size_<size>` node (`static_size`) brings it to Meta's exact pixels (1:1 at 1080×1080; a landscape size the same way). Every 9:16 Reels or Stories covers then passes a free `static_safe_fit` node (`safe_<size>`), which shrinks and shifts the picture just enough (to 85% at most) to bring a word or the logo a few points past the band back inside it; a layout further out passes unchanged and the run lays it out again. The 9:16 is never extended instead (the 4:5 set whole in the clear band, its scene continued to the frame) nor set over its own blurred copy: rated in October 2026, the first showed the ad small inside a continued scene ("text super tiny") and the second "looks bad". The master shape is one decision in the code (`DEFAULT_MASTER_SHAPE`: `relayout`; the `9x16_band` and `4x5_adapt` shapes, kept for the bench, make the sizes by cutting or reframing the master, and the person rating them rejected every such way; a canvas from the retired band master, which extended a square master into the 9:16, is upgraded to the re-layout graph on its next run). Placing the logo file on every size instead of having the model draw it (`logo_composite` nodes) is a scaffold option, off until it is proven on the bench.
 - Readers go by these roles, never by node ids: the run review, `--fix`, `--use-variant`, `--regenerate static_variants`, and the dashboard. A static ad scaffolded before this graph (`ad_a`/`ad_b`/`ad_c`, `pick`, re-lays `ad_4x5`/`ad_9x16`, `fit_4x5`) is **upgraded in place** on its next `validate` or `run`: prompt.json, every asset slot, the model and the formats are kept, the render part becomes this graph, and the creative is not reported as changed by the upgrade. Its old runs stay readable.
 - Detected elements are **never slots to fill**: the source's people, products, animals and badges are its world. The original image stays on the canvas as the record of the source; when `prompt.json` keeps its concept (`concept.verdict` `use` or `adapt`) the first renders are shown it **for its structure only**, and otherwise it never reaches a render. Only the people cast from a company Avatar and the brand's logo become pictures on the canvas; everything else is described in words (a picture of ours is added through a reference's `image`). A **person** with no avatar needs no slot and no casting: the render draws an invented, realistic person from the element's description. With `--avatar <handle>` (for an actor who recurs across ads) the slot ingests the Avatar's identity sheet and `prompt.json` describes that person in the Avatar's own words. A reference's `avatar` names who was cast and draws no face on its own: its face reaches the render only with a picture behind it (the `--avatar` slot, or the identity sheet URL from `baker avatars get <handle>` as the reference's `image`).
 
@@ -6556,6 +6556,7 @@ baker creatives publish ./canvas/<run_id>/final.mp4 --title "Spring Offer Reel"
 | `--slug <slug>` | Creative slug (`src/creatives/<slug>/`): attaches the image or video to that creative's row, marks it `published`. Required for a video. With `--run-id` an image publish also sets `status: published` in the creative's `_definition.md` (`definitionStatus` in the output). |
 | `--run-id <r_…>` | Pins the approved generation from the creative's run history as the published one. |
 | `--output <label\|slot\|n>` | With `--run-id` and no file: which of the run's images to save when it made several (a label such as `4:5`, a slot, or a 1-based number). |
+| `--approved` | The person's own message asked to save it. Needed only for an image run rendered in the current chat turn (`BAKER_TURN_ID`), which they have not seen yet; without it that save is refused. |
 | `--source-reference-url <url>` | Original reference ad URL, recorded on the creative. |
 | `--context <text>` | Optional describe context for the uploaded image or video. |
 

package/dist/{chunk-RFZYRG2Q.js → chunk-32UJLGZX.js}

@@ -1,10 +1,10 @@
 import {
   ApiError,
   apiGet
-} from "./chunk-Q6F5UNOU.js";
+} from "./chunk-UKHD423A.js";
 import {
   getEnv
-} from "./chunk-FQOKJYKN.js";
+} from "./chunk-PR3CXBAT.js";
 
 // src/error-handler.ts
 var REQUEST_CONNECTION_PLATFORM = {
@@ -298,4 +298,4 @@ export {
   writeAdsOutput,
   resolveCustomerId
 };
-//# sourceMappingURL=chunk-RFZYRG2Q.js.map
+//# sourceMappingURL=chunk-32UJLGZX.js.map

package/dist/{chunk-GCTR4MDE.js → chunk-5DJPTQU5.js}

@@ -1,6 +1,6 @@
 import {
   debugLogSetting
-} from "./chunk-FQOKJYKN.js";
+} from "./chunk-PR3CXBAT.js";
 
 // src/debugLog.ts
 import { appendFileSync, mkdirSync, renameSync, statSync } from "fs";
@@ -212,4 +212,4 @@ export {
   readBodyForLog,
   installStreamTaps
 };
-//# sourceMappingURL=chunk-GCTR4MDE.js.map
+//# sourceMappingURL=chunk-5DJPTQU5.js.map

package/dist/{chunk-GTHLEQHU.js → chunk-APTNJ2EY.js}

@@ -1,7 +1,7 @@
 import {
   debugLogHttp,
   readBodyForLog
-} from "./chunk-GCTR4MDE.js";
+} from "./chunk-5DJPTQU5.js";
 import {
   API_KEY_FORMAT_MESSAGE,
   DIRECT_ROUTE,
@@ -22,7 +22,7 @@ import {
   refuseNonPublicUrl,
   requireConnection,
   shouldEscalate
-} from "./chunk-FQOKJYKN.js";
+} from "./chunk-PR3CXBAT.js";
 
 // ../../.pnpm-store/v10/links/@/safe-stable-stringify/2.5.0/766ad107c42df28eeadaf0ac2f9089c8a9e200f6ba0b2b1a817b8d536cbdcc49/node_modules/safe-stable-stringify/index.js
 var require_safe_stable_stringify = __commonJS({
@@ -4999,10 +4999,10 @@ var GPT_IMAGE_BRIEF = {
   // (.context/static-review/BENCH.md#E3).
   // quality max: the deterministic checks tie (shippable 9-12% on every step, n=100), but on the
   // E4 gallery (30 rated pairs) the user's win rate is max 69%, xhigh 54%, high 47%, Flare 34%,
-  // and the judge's on the same pairs 73%, 65%, 39%, 31%. The user: results over cost ("use the
-  // best one, not the cheaper"), so the step is the best, at $0.166 per image vs $0.085 for xhigh
-  // and 80 s vs 44 s (BENCH.md#E4, § "E1 v2").
-  params: { n: 1, quality: "max", imageSize: "2K" }
+  // and the judge's on the same pairs 73%, 65%, 39%, 31%. Max costs $0.166 per image vs $0.085
+  // for xhigh and 80 s vs 44 s (BENCH.md#E4, § "E1 v2"); a round renders twice in a row (the
+  // options, then each size), so three ads took most of an hour. Oct 2026 the user chose xhigh.
+  params: { n: 1, quality: "xhigh", imageSize: "2K" }
 };
 var GEMINI_IMAGE_BRIEF = {
   // "Use Natural Language & Full Sentences… as if you were briefing a human artist", and
@@ -8383,14 +8383,16 @@ function scaleText(plan, spec, flowing, zoned, short) {
   const rows = hasRows(spec) ? flowing ? " Rows, columns and lists are restacked for the tall frame, one item per line, each larger." : " Rows, columns and lists are restacked for the tall frame, one item per line, each larger, rather than shrunk." : "";
   return flowing ? `The words are set for this frame, larger than in Image 1: the headline large enough to fill most of the band's width, and every line, the logo and the button spread through the band's height, inside its edges.${rows}` : `Scale: larger than in Image 1: the headline set large enough to fill most of the band's width; every line, the logo and the button spread through the band's height, inside its edges, never a small block in its middle.${rows}`;
 }
-var ZONE_HEADROOM = 0;
+var COVERED_HEADROOM = { top: 0.06, bottom: 0.08, side: 0.02 };
+var NO_HEADROOM = { top: 0, bottom: 0, side: 0 };
 var pct = (share) => `${Math.round(share * 100)}%`;
 function zonesText(plan, flowing, ground, short) {
-  const top = plan.safe.top + ZONE_HEADROOM;
-  const bottom = plan.safe.bottom + ZONE_HEADROOM;
-  const side = Math.max(plan.safe.left, plan.safe.right) + ZONE_HEADROOM;
-  const band = `from ${pct(top)} to ${pct(1 - bottom)} of the height, at least ${pct(side)} in from each side`;
   const covered = staticSafeArea(plan.format, plan.placements).placement !== "feed";
+  const room = isTall(plan) && covered ? COVERED_HEADROOM : NO_HEADROOM;
+  const top = plan.safe.top + room.top;
+  const bottom = plan.safe.bottom + room.bottom;
+  const side = Math.max(plan.safe.left, plan.safe.right) + room.side;
+  const band = `from ${pct(top)} to ${pct(1 - bottom)} of the height, at least ${pct(side)} in from each side`;
   if (short && !flowing) {
     const around = isTall(plan) && covered ? "only the ad's own picture, larger, never a flat colour, a panel or a square" : "only the ad's own picture";
     return `Zones: every word, the logo and the button sit ${band}; around them, ${around}.`;
@@ -8399,7 +8401,7 @@ function zonesText(plan, flowing, ground, short) {
     return flowing ? `The logo, every line and the button sit ${band}; around them the frame shows only the ad's own picture.` : `Zones: the logo, every line and the button sit ${band}; around them, only the ad's own picture.`;
   }
   const picture = ground ? "Image 1's background with its shapes and graphics enlarged in the brand's colours" : "Image 1's people, product and setting extended";
-  return flowing ? `The words' band runs ${band}, holding the logo, every line and the button. Above and below it the app's interface covers the frame, so there it shows only the ad's own picture, larger: ${picture}, edge to edge.` : `Zones: the logo, every line and the button sit in one band, ${band}; the top ${pct(top)} and the bottom ${pct(bottom)} carry neither words, the logo nor the button. Above and below it, where the app's interface covers the frame, only the ad's own picture, larger: ${picture}, never a flat empty colour, a panel or a square on a background.`;
+  return flowing ? `The words' band runs ${band}, holding the logo, every line and the button. Above and below it the app's interface covers the frame, so there it shows only the ad's own picture, larger: ${picture}, edge to edge.` : `Zones: the app's buttons and captions cover the top and the bottom of this frame, hiding any word there. The logo, every line and the button sit in one band, ${band}: the headline starts below the top fifth, and the button ends directly under the last line, never at the bottom edge where a feed ad puts it. Above and below it, where the app's interface covers the frame, only the ad's own picture, larger: ${picture}, never a flat empty colour, a panel or a square on a background.`;
 }
 function adaptBrief(spec, profile, opts) {
   const { plan, images } = opts;
@@ -8674,7 +8676,7 @@ function sizeNode(s, adaptFormats) {
   const index = adaptFormats.indexOf(s.adaptFormat ?? s.format);
   return magick(id, STATIC_SIZE_ROLE, `$ref:adapt.images#${index}`, fillArgs(s.size));
 }
-var RELAYOUT_QUALITY = "max";
+var RELAYOUT_QUALITY = "xhigh";
 function relayoutNodes(g) {
   return g.plan.sizes.filter((s) => s.from === "relayout").map((s) => relayoutNode(s, g));
 }
@@ -15995,6 +15997,12 @@ async function reviewRenderedStatic(opts) {
   const formats = measured.map((m) => m.format);
   const lines = opts.lines ?? opts.copy.map((text3) => ({ text: text3, lettering: "designed" }));
   const every = { lines, ...everySizeChecks(opts) };
+  const ask2 = (measuredRead) => {
+    const prompt = buildStaticReviewPrompt(reviewPromptOpts(opts, formats, measuredRead, reference !== null));
+    const shown = [...measuredRead ? stills : [...stills, ...thumbs], ...reference ? [reference] : []];
+    return askAboutStills(shown, prompt, opts.onSpend);
+  };
+  const asking = ask2(true);
   const checks = await Promise.all(
     originals.map(
       (bytes, i) => checkStaticImage(bytes, {
@@ -16006,9 +16014,8 @@ async function reviewRenderedStatic(opts) {
   );
   const read = checks.every((c) => c.status === "checked");
   const measuredText = read ? checks.flatMap((c) => c.status === "checked" ? c.findings : []) : [];
-  const prompt = buildStaticReviewPrompt(reviewPromptOpts(opts, formats, read, reference !== null));
-  const shown = [...read ? stills : [...stills, ...thumbs], ...reference ? [reference] : []];
-  const answer = await askAboutStills(shown, prompt, opts.onSpend) ?? await askAboutStills(shown, prompt, opts.onSpend);
+  const first = read ? await asking : (void asking.catch(() => null), null);
+  const answer = first ?? await ask2(read);
   if (answer === null) return unreviewed(shape, measuredText);
   const text2 = read ? measuredText : copyFindings(answer, opts.copy, formats, opts.scrolled);
   const boxesByFormat = read ? Object.fromEntries(checks.flatMap((c, i) => c.status === "checked" ? [[formats[i] ?? "", c.boxes]] : [])) : {};
@@ -18564,4 +18571,4 @@ export {
   defaultRegistry,
   createEngineFromEnv
 };
-//# sourceMappingURL=chunk-GTHLEQHU.js.map
+//# sourceMappingURL=chunk-APTNJ2EY.js.map