@milenyumai/film-kit 1.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,429 @@
1
+ ---
2
+ name: prompt-structure
3
+ description: Image and video prompt templates, camera movement vocabulary, stabilization styles, and prompt engineering best practices for Veo 3.1.
4
+ ---
5
+
6
+ # Prompt Structure & Engineering
7
+
8
+ > **Philosophy:** Precision breeds quality. Structure enables creativity.
9
+ > **Core Principle:** Every word must earn its place. No fluff. No ambiguity.
10
+
11
+ ---
12
+
13
+ ## 🎯 Core Principles
14
+
15
+ | Principle | Description |
16
+ |-----------|-------------|
17
+ | **Simplicity Over Complexity** | Start simple, iterate by adding detail |
18
+ | **Positive Framing** | Describe what SHOULD happen (negative prompts less effective) |
19
+ | **Specific Action Verbs** | "strides" > "walks", "glides" > "moves" |
20
+ | **Professional Terms** | Use industry-standard cinematography language |
21
+ | **Ideal Length** | 80-120 words per video prompt |
22
+ | **Reference Commands First** | Always start with reference image instructions |
23
+ | **Safety First** | Always consider filter implications |
24
+ | **Short Sentence Rule** | Split long sentences across shots |
25
+
26
+ ---
27
+
28
+ ## Image Prompt Structure
29
+
30
+ ### ✏️ Prompt Akış Sırası (Veo Optimizasyonu)
31
+
32
+ Veo ilk cümleye en çok ağırlık verir. **Her prompt bu sırayla yazılmalıdır:**
33
+
34
+ | Sıra | İçerik | Örnek |
35
+ |------|--------|-------|
36
+ | **1** | Tek cümle sahne özeti | "Cinematic close-up of a young soldier in a dim bunker." |
37
+ | **2** | Kim/Ne/Nerede | Karakter fiziksel tanımı, kıyafet, konum |
38
+ | **3** | Aksiyon | Ne yapıyor, mikro-davranış, oyunculuk |
39
+ | **4** | Kamera + Lens | "85mm f/2.0, shallow DOF, handheld" |
40
+ | **5** | Işık + Atmosfer | "Warm oil lamp, dramatic side shadows" |
41
+ | **6** | Audio direction | Sadece video prompt'larında |
42
+ | **7** | Avoid line | HER prompt'ta zorunlu |
43
+
44
+ > **İlk cümle = "Bu shot ne?" sorusunun tek cümlelik cevabı olmalı.**
45
+
46
+ ### Standard Template
47
+
48
+ ```
49
+ [REFERENCE LOCK section if applicable]
50
+
51
+ Cinematic still frame of [subject with reference adherence] in [frozen pose].
52
+ [Environment + time of day].
53
+ Lighting: [specific setup].
54
+ Camera: [framing], [lens mm], [aperture], photorealistic, crisp focus, no motion blur, no text.
55
+ [Safety injection if needed]
56
+
57
+ [Full avoid line]
58
+ ```
59
+
60
+ ### Example: Character in Environment
61
+
62
+ ```
63
+ [REFERENCE LOCK] Using uploaded reference image: The soldier - Match this person's face EXACTLY. Do NOT invent new facial features.
64
+
65
+ Cinematic still frame of the person from the reference, wearing dusty Ottoman artillery uniform, standing beside a massive artillery piece, one hand on the cannon barrel, determined expression. Ottoman coastal fortification, smoke-filled atmosphere, late afternoon harsh sunlight.
66
+ Lighting: Dramatic side-light from setting sun, long shadows, dust particles visible in light beams.
67
+ Camera: Medium shot, 35mm lens, f/4, photorealistic, crisp focus, no motion blur, no text.
68
+ Maintaining exact likeness from the reference throughout. Historical reenactment style. Documentary approach.
69
+
70
+ Avoid: blurry, low-res, noise, distorted faces, bad anatomy, extra limbs/fingers, plastic skin, waxy skin, airbrushed skin, on-screen text, watermark, logo, cartoon style, CGI look, different face than reference.
71
+ ```
72
+
73
+ ---
74
+
75
+ ## Video Prompt Structure
76
+
77
+ ### Standard Template
78
+
79
+ ```
80
+ Use the provided first frame as exact starting composition. End on provided last frame as exact final composition.
81
+
82
+ [REFERENCE LOCK section if applicable - maintaining character/object consistency from references]
83
+
84
+ [Subject + specific attributes] + [Action verb + motion details] + [Environment + time of day] + [Lighting details].
85
+ Camera: [start → end position], [movement type], [stabilization].
86
+
87
+ Audio direction:
88
+ - Language: [LANGUAGE]
89
+ - Type: [Type]
90
+ - Dialogue transcript: [Lines or NONE]
91
+ - SFX: [Effects list]
92
+ - Ambience: [Background sounds]
93
+ - Music: NONE
94
+ - Mix target: [Percentages]
95
+ - No on-screen subtitles/captions.
96
+
97
+ [Full avoid line]
98
+ ```
99
+
100
+ ### Example: Action Shot
101
+
102
+ ```
103
+ Use the provided first frame as exact starting composition. End on provided last frame as exact final composition.
104
+
105
+ [REFERENCE LOCK] Maintaining exact character likeness from the uploaded character reference throughout. The same person from the reference, not a similar one. Do NOT modify facial features.
106
+
107
+ The person from the reference, wearing strained Ottoman artillery uniform, hoists a massive artillery shell onto his shoulder. Visible effort and strain. Beside him, a younger soldier in matching uniform pushes from behind, providing support. Ottoman artillery position, smoke drifting through, harsh afternoon light casting dramatic shadows. Dust particles float in sunbeams.
108
+ Camera: Low angle medium shot, slight push-in on the soldiers' straining faces, handheld organic movement, capturing the gravity of their effort.
109
+
110
+ Audio direction:
111
+ - Language: TURKISH
112
+ - Type: Mixed
113
+ - Dialogue transcript: "Güç ve kuvvet sahibi Allah'tır. Haydi Bismillah."
114
+ - SFX: Heavy metal scraping, grunting with effort, fabric straining, boot scuffing on stone
115
+ - Ambience: Distant cannon fire, wind, smoke whooshing
116
+ - Music: NONE
117
+ - Mix target: Dialogue 40%, SFX 40%, Ambience 20%
118
+ - No on-screen subtitles/captions.
119
+
120
+ Avoid: distorted faces, morphing, bad anatomy, extra limbs/fingers, blurry, flickering, inconsistent lighting, unnatural motion, on-screen text, watermark, cartoon style, CGI motion, different face than reference.
121
+ ```
122
+
123
+ ---
124
+
125
+ ## Camera Movement Vocabulary
126
+
127
+ ### Primary Movements
128
+
129
+ | Movement | Description | Use When |
130
+ |----------|-------------|----------|
131
+ | **Dolly push-in** | Camera moves toward subject | Building intensity, focus |
132
+ | **Dolly pull-out** | Camera moves away from subject | Reveal, context, ending |
133
+ | **Pan left/right** | Camera rotates horizontally | Following action, scanning |
134
+ | **Tilt up/down** | Camera rotates vertically | Reveal height, power dynamics |
135
+ | **Crane rise** | Camera elevates vertically | Grand reveal, establishing |
136
+ | **Crane descend** | Camera lowers vertically | Intimate approach |
137
+ | **Orbit** | Camera circles subject | Dramatic emphasis, 360 view |
138
+ | **Tracking** | Camera follows alongside | Following movement |
139
+ | **Rack focus** | Focus shifts between planes | Attention shift, reveal |
140
+ | **Whip pan** | Camera rotates rapidly (snap) | Transition between subjects, hide cuts, energy burst |
141
+
142
+ ### Movement Combinations
143
+
144
+ ```
145
+ "Slow dolly push-in while panning left"
146
+ "Crane rise into tracking shot"
147
+ "Orbit at medium distance, maintaining focus on subject"
148
+ ```
149
+
150
+ ---
151
+
152
+ ## Stabilization Styles
153
+
154
+ | Style | Effect | Use For |
155
+ |-------|--------|---------|
156
+ | **Locked/Tripod** | No movement, static | Dialogue, still moments |
157
+ | **Gimbal smooth** | Stabilized glide | Premium, commercial feel |
158
+ | **Handheld organic** | Slight shake | Documentary, tension, realism |
159
+ | **Handheld intense** | Visible shake | Action, chaos, urgency |
160
+ | **Drone stabilized** | Smooth aerial | Establishing, landscapes |
161
+ | **Steadicam** | Smooth follow | Long takes, walking with character |
162
+
163
+ ---
164
+
165
+ ## Shot Duration Guidelines
166
+
167
+ | Scene Type | Duration | Why |
168
+ |------------|----------|-----|
169
+ | **Action/Impact** | 4s | Quick cuts maintain energy |
170
+ | **Standard scene** | 8s | Balanced, most common |
171
+ | **Emotional/drama** | 8s | Let moment breathe |
172
+ | **Dialogue heavy** | 8s | Time for delivery |
173
+ | **Establishing** | 8s | Set scene quickly |
174
+
175
+ ---
176
+
177
+ ## 🎲 Seed Parameter (Tutarlılık Kontrolü)
178
+
179
+ Veo, `seed` parametresiyle aynı prompt'tan **daha deterministik** sonuçlar üretebilir.
180
+
181
+ ### Ne Zaman Kullanılır?
182
+
183
+ | Durum | Seed Stratejisi |
184
+ |-------|----------------|
185
+ | **Aynı sahnede birden fazla açı** | Aynı seed → karakter/mekân tutarlılığı artar |
186
+ | **Coverage shot'ları (OTS A/B)** | Aynı seed → kıyafet/ışık/mekan sürekliliği |
187
+ | **Farklı sahneler** | Farklı seed → doğal varyasyon |
188
+ | **Yeniden üretim (re-render)** | Önceki seed → benzer sonuç |
189
+
190
+ ### Pratik Kullanım
191
+
192
+ ```
193
+ Seed AYNI tut:
194
+ - Aynı sahne, farklı kamera açısı
195
+ - Master shot + coverage arasında
196
+ - Extend ile uzatma öncesi/sonrası
197
+
198
+ Seed DEĞİŞTİR:
199
+ - Farklı lokasyon/zaman
200
+ - Tamamen yeni sahne
201
+ - Farklı karakter girişi
202
+ ```
203
+
204
+ ### Vertex AI Seed Desteği
205
+
206
+ - Destekleyen arayüzlerde seed parametresi gönderilebilir
207
+ - Aynı seed + aynı prompt = daha benzer (ama garanti değil) çıktı
208
+ - Referans görselle birlikte seed → tutarlılık daha da güçlenir
209
+
210
+ > **Not:** Seed tek başına tutarlılık garantisi vermez. **Referans kilitleme + seed** kombinasyonu en iyi sonucu verir.
211
+
212
+ ---
213
+
214
+ ## 🤖 Prompt Rewriter Davranışı (Veo İç Mekanizması)
215
+
216
+ Veo 3/3.1'de dahili bir **prompt enhancement/rewriter** mekanizması çalışır. Bu mekanizmanın nasıl çalıştığını bilmek, prompt kontrolünü artırır.
217
+
218
+ ### Nasıl Çalışır?
219
+
220
+ | Durum | Model Davranışı |
221
+ |-------|----------------|
222
+ | **Kısa prompt** (10-20 kelime) | Model agresif biçimde "tamamlar" → kontrol kaybı |
223
+ | **Orta prompt** (40-60 kelime) | Kısmi tamamlama → orta kontrol |
224
+ | **Uzun/spesifik prompt** (80-120 kelime) | Minimal tamamlama → yüksek kontrol |
225
+
226
+ ### Kontrol Stratejisi
227
+
228
+ ```
229
+ ❌ ZAYIF KONTROL (model çok tamamlar):
230
+ "Two soldiers talking in a bunker"
231
+
232
+ ✅ GÜÇLÜ KONTROL (model az tamamlar):
233
+ "Two Ottoman soldiers, early 30s, full dark mustaches, dusty khaki uniforms,
234
+ sitting at a wooden table inside a cramped stone bunker. Dim oil lamp light,
235
+ warm orange glow on faces, deep shadows on walls. Medium two-shot, 50mm, f/2.8,
236
+ handheld organic. Older soldier leans forward, speaking intently. Younger soldier
237
+ listens, eyes cast down, tension in his jaw."
238
+ ```
239
+
240
+ ### Kritik Kurallar
241
+
242
+ 1. **80-120 kelime hedefi** doğrudan rewriter'ın müdahale alanını daraltır
243
+ 2. **Kamera, ışık, davranış, ses** ne kadar detaylı → model o kadar az uydurur
244
+ 3. Veo 3/3.1'de rewriter'ı **kapatma seçeneği yoktur** — spesifik yazmak tek çözüm
245
+ 4. Kısa prompt'larda model genişlettiği prompt'u gösterir → kontrol ederek öğrenebilirsin
246
+
247
+ > **Altın Kural:** Prompt'taki her detay, model'in "yaratıcı boşluğunu" daraltır. Daha fazla detay = daha fazla kontrol.
248
+
249
+ ---
250
+
251
+ ## Lens Selection Guide
252
+
253
+ | Lens | Effect | Use For |
254
+ |------|--------|---------|
255
+ | **24mm** | Wide, environmental | Establishing, scale |
256
+ | **35mm** | Natural wide | Groups, environment + subject |
257
+ | **50mm** | Natural, most realistic | Standard coverage |
258
+ | **85mm** | Portrait, compression | Close-ups, intimacy |
259
+ | **135mm** | Telephoto, compression | Distant observation, isolation |
260
+ | **200mm+** | Extreme compression | Surveillance, extreme distance |
261
+
262
+ ### Aperture Effects
263
+
264
+ | Aperture | Depth of Field | Use For |
265
+ |----------|----------------|---------|
266
+ | **f/1.4-2.0** | Very shallow | Extreme focus isolation |
267
+ | **f/2.8** | Shallow | Portrait, subject isolation |
268
+ | **f/4** | Medium | Balanced, some background |
269
+ | **f/5.6-8** | Deep | Environmental, context |
270
+ | **f/11+** | Very deep | Landscapes, everything sharp |
271
+
272
+ ---
273
+
274
+ ## Action Verb Upgrades
275
+
276
+ Replace weak verbs with specific, cinematic alternatives:
277
+
278
+ | Weak | Strong Alternatives |
279
+ |------|---------------------|
280
+ | walks | strides, trudges, marches, staggers, shuffles |
281
+ | runs | sprints, dashes, bolts, charges, scrambles |
282
+ | looks | gazes, peers, glares, scans, scrutinizes |
283
+ | talks | whispers, shouts, mutters, commands, pleads |
284
+ | moves | glides, lurches, pivots, lunges, retreats |
285
+ | holds | grips, clutches, cradles, hefts, brandishes |
286
+ | falls | collapses, tumbles, crumbles, drops, plummets |
287
+
288
+ ---
289
+
290
+ ## Cinematic Description Templates
291
+
292
+ ### Character Introduction
293
+
294
+ ```
295
+ [Character description], [age], [distinctive feature], [costume].
296
+ [Posture/body language] communicating [emotion/intent].
297
+ [Specific detail] catching the light.
298
+ ```
299
+
300
+ ### Action Moment
301
+
302
+ ```
303
+ [Character] [strong action verb] with [quality of motion].
304
+ [Physical detail] visible as [evidence of effort/emotion].
305
+ [Environmental reaction] to the action.
306
+ ```
307
+
308
+ ### Emotional Beat
309
+
310
+ ```
311
+ [Character's] expression shifts from [state A] to [state B].
312
+ [Micro-detail: eyes, mouth, hands] betraying [inner state].
313
+ A beat of [silence/stillness/tension] before [next action].
314
+ ```
315
+
316
+ ### Environmental Establishing
317
+
318
+ ```
319
+ [Wide view] of [location] at [time of day].
320
+ [Atmospheric elements: light, weather, particles] defining the mood.
321
+ [Scale elements] establishing the [grandeur/intimacy/danger].
322
+ ```
323
+
324
+ ---
325
+
326
+ ## Prompt Length Guidelines
327
+
328
+ | Prompt Type | Words | Structure |
329
+ |-------------|-------|-----------|
330
+ | **Image prompt** | 60-100 | Reference + Description + Technical + Avoid |
331
+ | **Video prompt** | 80-120 | Frame instructions + Reference + Action + Camera + Audio + Avoid |
332
+
333
+ ### If Prompt Too Long
334
+
335
+ 1. Split into multiple shots
336
+ 2. Remove redundant descriptors
337
+ 3. Combine phrases ("harsh afternoon light" not "harsh light in the afternoon")
338
+ 4. Use professional shorthand ("CU" for close-up in notes, full term in prompt)
339
+
340
+ ---
341
+
342
+ ## Quality Checklist
343
+
344
+ Before finalizing any prompt:
345
+
346
+ - [ ] Reference section at start? (if applicable)
347
+ - [ ] Subject clearly described?
348
+ - [ ] Action verbs specific?
349
+ - [ ] Camera movement defined?
350
+ - [ ] Lens and aperture specified?
351
+ - [ ] Lighting described?
352
+ - [ ] Audio direction complete? (video only)
353
+ - [ ] Avoid line included?
354
+ - [ ] Length within guidelines?
355
+
356
+ ---
357
+
358
+ ## 🔄 Re-Take Strategy (Yeniden Üretim Protokolü)
359
+
360
+ Veo glitch veya istenmeyen sonuç ürettiğinde:
361
+
362
+ ### Hata Tipine Göre Müdahale
363
+
364
+ | Hata | Prompt'ta Ne Değiştir | Seed |
365
+ |------|----------------------|------|
366
+ | **Yüz bozulması** | "photorealistic face, sharp features" ekle, Avoid'a "distorted face, morphing" güçlendir | Yeni seed dene |
367
+ | **Ekstra uzuv/parmak** | Avoid'a "extra limbs, extra fingers, merged hands" ekle | Yeni seed |
368
+ | **Hareket tutarsızlığı** | Hareketi sadeleştir, daha az eylem | Aynı seed, sadeleşmiş prompt |
369
+ | **Işık zıplaması** | Işık kaynağını daha spesifik tanımla | Aynı seed |
370
+ | **Karakter tutarsız** | Reference lock'u güçlendir, "EXACTLY" 3x tekrarla | Aynı seed |
371
+ | **Arka plan kayması** | Arka planı daha detaylı tanımla | Aynı seed |
372
+ | **Robotik hareket** | "Organic natural movement" ekle, micro-davranış koy | Yeni seed |
373
+
374
+ ### Re-Take Süreci
375
+
376
+ ```
377
+ 1. İLK DENEME: Orijinal prompt
378
+ 2. Glitch varsa: Hata tipini belirle
379
+ 3. İKİNCİ DENEME: Prompt ayarla (yukarıdaki tabloya göre)
380
+ 4. Hâlâ sorun: Seed değiştir + prompt'u sadeleştir
381
+ 5. ÜÇÜNCÜ DENEME: Kamera açısını veya frame'i değiştir
382
+ 6. 3 denemeden sonra: Shot'u daha kısa parçalara böl (8s→4s+4s)
383
+ ```
384
+
385
+ > **Kural:** 3 başarısız denemeden sonra shot'u bölmek genellikle en etkili çözümdür.
386
+
387
+ ---
388
+
389
+ ## 📏 Coverage Prompt Yazım Standartları
390
+
391
+ > **❗ COVERAGE = ANA SHOT KALİTESİNDE OLMALIDIR. KISA PROMPT YASAKTIR.**
392
+
393
+ ### Minimum Gereksinimler
394
+
395
+ | Prompt Tipi | Min Kelime | İçermesi Gerekenler |
396
+ |-------------|-----------|---------------------|
397
+ | Coverage Image | **60 kelime** | Subject + camera + lens + lighting + emotion + Avoid |
398
+ | Coverage Video | **60 kelime** | Action + camera movement + atmosphere + Audio block + Avoid |
399
+
400
+ ### Kısa Coverage Nasıl Uzatılır?
401
+
402
+ ```
403
+ ❌ KISA (25 kelime — REDDEDİLECEK):
404
+ "Close-up reaction shot of the soldier listening. Soft light.
405
+ Avoid: blurry, distorted."
406
+
407
+ ✅ TAM (70 kelime — KABUL):
408
+ "Close-up reaction shot of a young soldier, early 20s, thick dark mustache,
409
+ dusty khaki uniform. Eyes fixed on the speaker, slight tension in jaw,
410
+ brow furrowed. Soft key light from screen-left, warm orange oil lamp glow.
411
+ 85mm f/2.0, shallow depth of field, background softly blurred.
412
+ Static camera, handheld organic micro-movement.
413
+ Photorealistic, cinematic grain.
414
+
415
+ Avoid: blurry, low-res, noise, distorted faces, bad anatomy, extra limbs/fingers,
416
+ plastic skin, waxy skin, on-screen text, watermark, logo, cartoon style, CGI look."
417
+ ```
418
+
419
+ ### Coverage'da Eklenmesi Gereken Detaylar
420
+
421
+ 1. **Karakter fiziksel tanımı** (yaş, saç, bıyık, kıyafet)
422
+ 2. **Duygusal durum** (gerilim, korku, merak — mikro-davranış)
423
+ 3. **Kamera** (lens mm, aperture, stabilization)
424
+ 4. **Işık** (yön, renk sıcaklığı, kaynak)
425
+ 5. **Avoid satırı** (tam versiyon, kısaltma YOK)
426
+
427
+ ---
428
+
429
+ > **Remember:** A great prompt is like a great shot list — nothing wasted, everything intentional. Write like a director who knows exactly what they want to see.