till/managarten
1
0
Fork 0
mirror of https://github.com/Memo-2023/mana-monorepo.git synced 2026-05-14 19:41:09 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

fix(wardrobe): prompt Ganzkörperfoto instead of Portrait for full try-on

Browse source 
The non-accessory Try-On prompts started with "Fotorealistisches
Portrait von mir", and both gpt-image-1 and gpt-image-2 read that as
a photographic framing hint ("head-and-shoulders crop") rather than a
general "picture of a person". Result: even with body-ref supplied
and a 1024×1536 portrait canvas, the model rendered a headshot and
ignored the body reference.

Keep "Portrait" only for the accessory path (brille/schmuck/hut) —
there the tight head framing is what we actually want for legibility.
Full-garment/outfit paths now say "Fotorealistisches Ganzkörperfoto
von mir im/in <garment> … stehend, von Kopf bis Fuß sichtbar" which
reliably biases the model to full-length framing.

Applies to both runGarmentTryOn (single-garment) and runOutfitTryOn
(outfit) — they share the same framing mistake.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Till JS 2026-04-24 13:56:29 +02:00
parent 95e85bdffd
commit 0cebb2411e
1 changed files with 11 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

13
apps/mana/apps/web/src/lib/modules/wardrobe/api/try-on.ts
View file

@ -130,10 +130,16 @@ export function isAccessoryOnlyOutfit(garments: Garment[]): boolean {
function composeDefaultPrompt(outfit: Outfit, accessoryOnly: boolean): string {
if (accessoryOnly) {
// Portrait is the right framing for accessories — we want a tight
// head-and-shoulders shot so the Brille/Schmuck/Hut is legible.
return `Fotorealistisches Portrait von mir mit ${outfit.name}, frontal, studio-Licht, neutraler Hintergrund, Fokus auf dem Accessoire`;
}
// Explicitly ask for a full-body frame — gpt-image-1/2 otherwise read
// "Portrait" as "headshot" and crop to head-and-shoulders, ignoring
// the body-ref. "Ganzkörperfoto … stehend, von Kopf bis Fuß sichtbar"
// is the German idiom that reliably biases the model to full-length.
const occasionHint = outfit.occasion ? ` (Anlass: ${outfit.occasion})` : '';
return `Fotorealistisches Portrait von mir im Outfit ${outfit.name}${occasionHint}, natürliches Licht, neutraler Hintergrund`;
return `Fotorealistisches Ganzkörperfoto von mir im Outfit ${outfit.name}${occasionHint}, stehend, von Kopf bis Fuß sichtbar, natürliches Licht, neutraler Hintergrund`;
}
export async function runOutfitTryOn(params: RunOutfitTryOnParams): Promise<RunTryOnResult> {
@ -237,9 +243,12 @@ export function isAccessoryGarment(garment: Garment): boolean {
function composeGarmentPrompt(garment: Garment, accessoryOnly: boolean): string {
if (accessoryOnly) {
// Headshot framing for face-only categories (Brille/Schmuck/Hut).
return `Fotorealistisches Portrait von mir mit ${garment.name}, frontal, studio-Licht, neutraler Hintergrund, Fokus auf dem Accessoire`;
}
return `Fotorealistisches Portrait von mir im/in ${garment.name}, natürliches Licht, neutraler Hintergrund`;
// Explicit full-body framing — see composeDefaultPrompt for the
// rationale. "Portrait" biases to headshot, "Ganzkörperfoto" doesn't.
return `Fotorealistisches Ganzkörperfoto von mir im/in ${garment.name}, stehend, von Kopf bis Fuß sichtbar, natürliches Licht, neutraler Hintergrund`;
}
/**