Add MiniMax and PinkCherry LTX as pickable video engines, with still or text start.

Shot 1 can run text-to-video; later shots still extend from the last frame. Recommend output is also normalized into shot-script cards.

Co-authored-by: Cursor <cursoragent@cursor.com>

utils/parseRecommend.ts

@@ -1,3 +1,5 @@
+import { parseShotScript } from './parseShots'
+
 export function parseEditPasses(text: string): string[] | null {
   const src = String(text || '').replace(/\r\n/g, '\n')
   const marker = /^[ \t]*---pass\s+(\d+)---[ \t]*$/gim
@@ -21,3 +23,40 @@ export function parseEditPasses(text: string): string[] | null {
 export function looksLikeShotScript(text: string) {
   return /^[ \t]*shot\s+[2-9]\d*\b/im.test(String(text || ''))
 }
+
+function stripFences(text: string) {
+  return String(text || '')
+    .replace(/^\s*```(?:[\w-]+)?\s*\n?/, '')
+    .replace(/\n?```\s*$/, '')
+    .trim()
+}
+
+function splitInlineShotMarkers(text: string) {
+  return String(text || '')
+    .replace(/\r\n/g, '\n')
+    .replace(/(^|[^A-Za-z])shot\s+(\d+)\s*[:.)\-]?\s*/gi, '$1\n\nshot $2\n')
+    .replace(/\n{3,}/g, '\n\n')
+    .trim()
+}
+
+function formatShotBody(body: string) {
+  let text = String(body || '').trim()
+  text = text.replace(/^\s*[:.\-–)]\s*/, '')
+  text = text.replace(/\s*\[SHOT CONFIGURATION\]\s*/gi, '\n\n[SHOT CONFIGURATION]\n')
+  text = text.replace(/\s*\[SUBJECT DIRECTION(?:\s*&\s*ACTION)?\]\s*/gi, '\n\n[SUBJECT DIRECTION & ACTION]\n')
+  text = text.replace(/\s*\[AUDIO CUES\]\s*/gi, '\n\n[AUDIO CUES]\n')
+  text = text.replace(/[ \t]+\n/g, '\n').replace(/\n{3,}/g, '\n\n').trim()
+  return text
+}
+
+export function normalizeShotScript(text: string) {
+  const src = splitInlineShotMarkers(stripFences(text))
+  const shots = parseShotScript(src)
+  if (!shots.length) return src
+  if (shots.length === 1 && !/^[ \t]*shot\s+\d+\b/im.test(src)) {
+    return formatShotBody(shots[0].prompt)
+  }
+  return shots
+    .map(shot => `shot ${shot.n}\n${formatShotBody(shot.prompt)}`.trim())
+    .join('\n\n')
+}