Update to Frameo.AI

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -1,28 +1,38 @@
 ---
 name: thumbnail-set
-description: "Three to five thumbnail concepts for a video: a face or product, a short title treatment, high contrast, delivered at upload size."
+description: "Three to five thumbnail concepts for a video, then variants of the chosen one: a face or product, a short title treatment, high contrast, delivered at upload size. Use for YouTube thumbnails, video covers and A/B test sets."
 ---
 
-Needs: the video's title; a face photo or the product photos; the platform (YouTube 16:9, Instagram 1:1 or 4:5)
-Credits: about 110–180 (3–5 images at ~36 each, the default model's 4K tier)
-Time: 10 minutes
-
 ## When
 
-After the video exists: a set to A/B. Not for the video itself.
+A set of thumbnail or cover options for a video, to pick from or A/B test. The video itself
+comes from `script-to-video`, `faceless-story-video` or `ugc-video`, not from here. A 9:16 Reels
+or Shorts cover for a product follows `get_skill_file(skill_id="thumbnail-set", path="references/reel-cover.md")`.
 
 ## Ask first
 
-1. **Title** — the 2–4 words that go on the image.
-2. **Subject** — a face (`show_upload` for files on the user's device, `import_media_url` for a web link; images attached to the chat do not reach Frameo, and `create_upload_url` is for clients that send the file themselves) or the product photos.
-3. **Platform** — sets the frame: YouTube 16:9, Instagram 1:1; for Instagram 4:5 generate at
-   3:4 (the nearest supported ratio) and crop at the end.
+All in one turn:
+
+1. **Title** — 2–6 words for the overlay, or none. It is burned in at step 6, never generated.
+2. **Who is in frame** — "Do you want yourself or a specific person in the thumbnail? Send a
+   face photo and the thumbnail keeps that exact face; otherwise say whether you want a generated
+   person or no person at all." Photos come in through `show_upload` for files on the user's device, `import_media_url` for a web link; images attached to the chat reach Frameo only through `import_attachments`, in apps that pass them such as ChatGPT, and `create_upload_url` is for clients that send the file themselves. A generated person only when the
+   user asks for one; never an invented real-looking person by default. No person means a product,
+   place, map or graphic concept.
+3. **Platform** — sets the frame: YouTube 16:9, Shorts 9:16, Instagram 1:1; for Instagram 4:5
+   generate at 3:4 (the nearest supported ratio) and crop at the end.
+4. **A reference thumbnail** (optional) — an example the user likes. Ask once whether to match it
+   closely or take it as loose inspiration. It is studied by eye only, for its concept, pose,
+   composition, background and emotion, and is never passed to a generate call.
+5. **How many** — one thumbnail, or a set of about four variants of the chosen concept (each a
+   different emotion or camera take, at most 16).
 
 ## Steps
 
 **Before any paid step.** The project: `list_projects` (or `create_project`) gives the `project_id`
 and, when the project has several modules, the `module_id`; every call below that takes a project —
-`estimate_cost` included — gets that same pair; without it those tools answer `project_needed`. The
+`estimate_cost` included — gets that same pair; without the project those tools answer `project_needed`, and without the
+module of a project that has several, `module_needed`, which lists its modules. The
 quote: one `estimate_cost(items=[…])` prices a stage in one call, with the same project, model,
 size and number of `image_urls` or `reference_image_urls` (`reference_count`) as each generate call, and returns a `quote_id` per item plus the total. Each generate call
 then passes its own item's `quote_id` and `confirmed_by_user=true`. A quote is single-use and lasts
@@ -30,7 +40,7 @@
 comes to more than the user approved is asked about again first. Generate calls return
 `generation_ids`; `wait_task` returns the links, and `show_generations` shows each stage's running
 and finished work in one card where the chat app displays Frameo cards: all the stage's ids at
-once, before its first `wait_task`, and the finished result with `final: true`.
+once, before its first `wait_task`; the card updates itself as the work finishes, so the same ids are not shown again.
 
 **The plan first.** Before the first paid call, the plan goes to the user in the chat as plain
 text: what will be made, in order, one line per generation (for a script, the shot list; for a
@@ -40,31 +50,59 @@
 generated until the user says yes. The user can drop or change lines; a changed line is priced
 again.
 
-**1. Concepts.** Propose 3–5 in one line each (big face + reaction, product + number, before/
-after, a question, a bold object on colour); the user picks or edits.
-
-**2. Images (~36 each at 4K).** `generate_image` per concept, each with its own quote,
-`image_urls=[face or product]`,
-prompt = the concept plus "high contrast, one focal point, space on the left third for
-text, no small details", `aspect_ratio` from Ask first (1:1, 16:9, 9:16 or 3:4) and
-`image_resolution` at the highest tier the model offers (`list_models(kind="image")`
-names them, e.g. 4K on the default; the quote is exact for that model and tier). Leave the title off the image: the exact title is burned in at step 3, not generated.
+**1. Concepts.** Read `get_skill_file(skill_id="thumbnail-set", path="references/frameworks.md")`. Every concept opens an
+information gap (the image raises a question the video answers) and stays true to the video: it
+may exaggerate, never promise what the video does not have. Propose 3–5 concepts in one line
+each, drawn from different frameworks (two can combine); the user picks or edits.
+
+**2. Prompts.** One prompt per variant, built in the order set out in
+`get_skill_file(skill_id="thumbnail-set", path="references/prompt-order.md")`: frame, scene, no text, subject, key elements,
+location, composition, background, lighting, grade. The subject is large, 40–60% of the frame,
+on a power third; the title takes the free quarter that composition leaves, so no fixed text
+area is asked for in the prompt. With a face photo, the identity-lock paragraph from that file
+goes in for each person, and every person gets the lighting rig. A split frame only when the
+user asks for a split, before/after or versus layout. Variants differ only by the emotion
+phrase or one camera-take line from that file.
+With no person in the concept (a landscape, a map, a product alone), variants differ by one camera
+take (wider, closer, lower) or one change of time of day or light instead.
+
+**3. Images (~36 each at 4K).** `generate_image` per variant, each with its own quote,
+`image_urls` = the face photos first (in character order), then any product photo or logo; none
+for a concept without them. `aspect_ratio` from Ask first, the default `image_model`, and
+`image_resolution` at the highest tier the model offers (`list_models(kind="image")` names them,
+e.g. 4K on the default; the quote is exact for that model and tier). No text in the image.
 `wait_task`, then show the set.
 
-**3. Title and exact size (free).** One `run_ffmpeg` per pick: `scale` + `crop` to the
-platform's pixel size (1280×720, 1080×1080, 1080×1350), and the title burned in from an `.ass`
-sidecar with the `ass` filter (Noto Sans, large and bold, thick outline, in the free space) unless the
-user prefers their own editor; ask which. `wait_task` for each
-output link.
+**4. Check (free).** Each image against the plan: `view_image(url, resolution="high")` shows
+whether the face matches its photo, whether any stray text or watermark slipped in, and whether
+hands and the hero element are clean. The small preview `wait_task` returns stands in for the
+feed: a thumbnail competes at about 120 px wide, so the emotion and the hero element must read
+at that size; if they do not, the composition is wrong, not the resolution. A failed image is
+made again once with the same prompt; one still failing is shown with the flaw named.
+
+**5. Tweaks (when asked).** One change at a time on the picked image: `generate_image` with
+`image_urls=[the picked image]` and a prompt that changes only that one thing (the expression,
+the background, its colour, the rim-light colour) and keeps everything else exactly as it is;
+the wording is in `get_skill_file(skill_id="thumbnail-set", path="references/prompt-order.md")`. Each tweak is quoted like
+any image, and its result becomes the picked image for the next one.
+
+**6. Title and exact size (free).** Ask whether the title is burned in or left for the user's
+own editor. One `run_ffmpeg` per pick: `scale` + `crop` to the platform's pixel size (1280×720,
+1080×1920, 1080×1080, 1080×1350), and the title burned in from an `.ass` sidecar with the `ass`
+filter, in one of the styles in `get_skill_file(skill_id="thumbnail-set", path="references/title-styles.md")`, placed in the
+free quarter of that image (looked at first) and never over a face. `view_image` on the result
+checks the title reads correctly. `wait_task` for each output link.
 
-**4. Deliver.** Links and `open_in_frameo`.
+**7. Deliver.** Links and `open_in_frameo`.
 
 ## Done
 
-The set on the canvas in one row.
-`open_in_frameo` on the result opens the project in the Frameo app, where every take sits on
-the canvas, ready for retakes and edits.
+The set in the Frameo project, kept together, each with its concept and, when made, its titled version.
+`open_in_frameo` on the result opens the project in the Frameo app, where every take is ready for retakes and edits.
 
 ## Files
 
-None.
+- `references/frameworks.md` — 16 concept frameworks, the information gap and the truth rule, read at step 1.
+- `references/prompt-order.md` — the prompt order, identity lock, emotions, camera takes, lighting rig, split frames and tweak wording, read at steps 2 and 5.
+- `references/title-styles.md` — five `.ass` title styles and where the title goes, read at step 6.
+- `references/reel-cover.md` — a 9:16 cover with the product low and the top third kept for the title.