Home · Skills · People & hiring

muapi-seedance-2

Expert Cinema Director skill for Seedance 2.0 (ByteDance) — high-fidelity video generation across Chinese, Global, and VIP tiers.

Paste into Claude, ChatGPT or Cursor.

Read the source632 lines

muapi-seedance-2 /

SKILL.md only. Everything else in the repo is on GitHub.

muapi-seedance-2/SKILL.md632 lines26.5 KBpushed 100d agoRawView on GitHub

1	---
2	slug: muapi-seedance-2
3	name: muapi-seedance-2
4	version: "0.3.0"
5	description: Expert Cinema Director skill for Seedance 2.0 (ByteDance) — high-fidelity video generation across Chinese, Global, and VIP tiers. Supports text-to-video, image-to-video, first-last-frame, omni reference, character training, omni-reference training, video editing, and watermark removal.
6	acceptLicenseTerms: true
7	---
8
9	# 🎬 Seedance 2.0 Cinema Expert
10
11	The definitive skill for "Director-Level" AI video orchestration.
12	Seedance 2.0 is not a descriptive model; it is an instructional model. It responds best to technical cinematography, physics directives, and precise camera grammar.
13
14	## Core Competencies
15
16	1. Text-to-Video (t2v): Generate cinematic video from a Director Brief — Chinese, Global, or VIP tier.
17	2. Image-to-Video (i2v): Animate 1–9 reference images — Chinese, Global (smart mode), or VIP tier.
18	3. Video Extension (extend): Seamlessly continue an existing Seedance 2.0 video (Chinese tier).
19	4. First & Last Frame (first-last): Interpolate a fluid video between a start image and end image (Global/VIP).
20	5. Omni Reference (omni): Full multimodal reference with images + audio + character refs (all tiers).
21	6. Omni Reference Training (omni-train): Train a custom persistent character for identity-consistent generation.
22	7. Character Sheet (character): Build a reusable character from 1–3 images (Chinese tier).
23	8. Video Edit (video-edit): Edit an existing video with a prompt + optional reference images (Chinese tier).
24	9. Watermark Removal (watermark-remove): Strip Seedance 2.0 watermarks (basic or Pro).
25
26	---
27
28	## 🏷️ Tiers
29
30	\| Tier \| Flag \| Censorship \| Aspect Ratios \| Duration \| Quality param \|
31	\|:---\|:---\|:---\|:---\|:---\|:---\|
32	\| Chinese (default) \| `--tier chinese` \| Low \| 16:9, 9:16, 4:3, 3:4 \| 5 / 10 / 15 s \| Yes (basic/high) \|
33	\| Global \| `--tier global` \| Standard \| + 21:9, 1:1 \| Any 4–15 s \| No \|
34	\| VIP \| `--tier vip` \| Low \| + 21:9, 1:1 \| Any 4–15 s \| No \|
35
36	Add `--fast` to any Global or VIP call to use the fast-queue variant (lower latency, same quality).
37
38	---
39
40	## 📥 Input Limits
41
42	\| Input Type \| Chinese i2v/omni \| Global/VIP i2v/omni \| Formats \| Max Size \|
43	\|:---\|:---\|:---\|:---\|:---\|
44	\| Images \| ≤ 9 \| ≤ 9 \| jpeg, png, webp \| 30 MB each \|
45	\| Videos \| ≤ 3 (omni only) \| Not supported \| mp4, mov \| 50 MB each \|
46	\| Audio \| ≤ 3 \| ≤ 3 \| mp3, wav \| 15 MB each \|
47	\| First-Last \| — \| 1–2 images \| jpeg, png, webp \| 30 MB each \|
48	\| Video Edit \| 1 video + ≤ 9 imgs \| — \| mp4 ≤ 10 MB / 15s \| — \|
49
50	Output: 4–15 seconds, auto-generated sound, 480p–720p.
51
52	---
53
54	## ⚠️ Restrictions
55
56	- No realistic human faces in uploaded images/videos (except character/omni-train modes).
57	- `--mode extend` requires a `request_id` from a prior `seedance-v2.0-t2v` or `seedance-v2.0-i2v` job.
58	- `--mode first-last` requires `--tier global` or `--tier vip`.
59	- Global/VIP omni does not support video references (images + audio only).
60	- `--quality` applies to Chinese tier only.
61
62	---
63
64	## 🔗 Core Syntax: The @ Reference System
65
66	Assign explicit roles to each uploaded asset. Tags differ by mode.
67
68	### Chinese Tier (i2v, omni)
69	```
70	@image1 @image2 ... @image9 (images_list order)
71	@video1 @video2 @video3 (video_files order)
72	@audio1 @audio2 @audio3 (audio_files order)
73	```
74
75	### Global/VIP Omni (omni-reference-no-video / vip-omni-reference)
76	```
77	@image1 @image2 ... @image9 (images_list order)
78	@audio1 @audio2 @audio3 (audio_files order)
79	```
80
81	### Character References (all tiers)
82	```
83	@character:<request_id> — from seedance-2-character or completed t2v/i2v job
84	@omni-character:<character_id> — from seedance-2-omni-reference-train output
85	```
86
87	### Role Assignment Table
88
89	\| Purpose \| Example Syntax \|
90	\|:---\|:---\|
91	\| First frame \| `@Image1 as the first frame` \|
92	\| Last frame \| `@Image2 as the last frame` \|
93	\| Character appearance \| `@Image1's character as the subject` \|
94	\| Scene / background \| `scene references @Image3` \|
95	\| Camera movement \| `reference @Video1's camera movement` \|
96	\| Action / motion \| `reference @Video1's action choreography` \|
97	\| Visual effects \| `completely reference @Video1's effects and transitions` \|
98	\| Rhythm / tempo \| `video rhythm references @Video1` \|
99	\| Voice / tone \| `narration voice references @Video1` \|
100	\| Background music \| `BGM references @Audio1` \|
101	\| Sound effects \| `sound effects reference @Video3's audio` \|
102	\| Outfit / clothing \| `wearing the outfit from @Image2` \|
103	\| Product appearance \| `product details reference @Image3` \|
104
105	### Multi-Reference Combination
106	```
107	@Image1's character as the subject, reference @Video1's camera movement
108	and action choreography, BGM references @Audio1, scene references @Image2
109	```
110
111	---
112
113	## 🏗️ Technical Specification: The Director Brief
114
115	Structure prompts using this six-component hierarchy. Order matters — composition first, texture and micro-motion last:
116
117	\| Component \| Instruction Type \| Example \|
118	\|:---\|:---\|:---\|
119	\| Scene \| Environment + Lighting \| "A rain-soaked cyberpunk street, magenta neon reflections on wet asphalt." \|
120	\| Subject \| Identity + Detail \| "A woman in a black trenchcoat, determined focus, cinematic skin textures." \|
121	\| Action \| Fluid Interaction \| "Walking forward through the crowd, coat billowing slightly in the wind." \|
122	\| Camera \| Movement + Lens + Speed \| "Medium tracking shot, 35mm lens, slow dolly backward over 6s. Subtle handheld jitter." \|
123	\| Audio \| Music + SFX + Ambience \| "Low ambient hum, distant traffic, single piano note at 5s. No dialogue." \|
124	\| Pacing/Style \| Timing + Mood + Grade \| "Cinematic epic, warm color grade, shallow DOF. Slow build — single action only, no scene cuts." \|
125
126	> Seedance 2.0 generates audio natively. Always include an Audio directive — even one sentence. Without it the model generates random ambient sound that may not match your scene.
127
128	### Time-Segmented Prompts (Recommended for 10s+ videos)
129	Break prompts into timed segments for precise control:
130	```
131	0–3s: [opening scene, camera move, establishing action]
132	3–6s: [mid-section development, subject in motion]
133	6–10s: [climax or key action beat]
134	10–15s: [resolution, brand/product hold, text/tagline fade in]
135	```
136
137	> Single-beat rule: Each segment should contain one action. 4–7s = one beat. 10–15s = 3–4 beats maximum. Overloading a segment with multiple narrative changes degrades output quality.
138
139	### Negative Prompting
140
141	Seedance 2.0 supports appending negative guidance directly in the prompt. Use plain language at the end:
142
143	```
144	[your director brief above]
145	Avoid: camera shake, jump cuts, lens distortion, overexposure, watermarks, text overlays.
146	```
147
148	Common negative additions:
149	- `Avoid: abrupt cuts, scene changes, multiple locations.` (for single-take shots)
150	- `Avoid: human faces, realistic people.` (for product-only content)
151	- `Avoid: fast motion, blur, unstable framing.` (for smooth product reveals)
152
153	---
154
155	## 🎥 Camera Language Reference
156
157	### Basic Movements
158	\| Term \| Description \|
159	\|:---\|:---\|
160	\| Push in / Slow push \| Camera moves toward subject \|
161	\| Pull back / Pull away \| Camera moves away from subject \|
162	\| Pan left/right \| Camera rotates horizontally \|
163	\| Tilt up/down \| Camera rotates vertically \|
164	\| Track / Follow shot \| Camera follows subject movement \|
165	\| Orbit / Revolve \| Camera circles around subject \|
166	\| One-take / Oner \| Continuous shot with no cuts \|
167
168	### Advanced Techniques
169	\| Term \| Description \|
170	\|:---\|:---\|
171	\| Hitchcock zoom (dolly zoom) \| Push in + zoom out — creates vertigo effect \|
172	\| Fisheye lens \| Ultra-wide distorted lens \|
173	\| Low angle / High angle \| Camera below/above subject \|
174	\| Bird's eye / Overhead \| Top-down view \|
175	\| First-person POV (FPV) \| Immersive subjective camera from character/object's eyes — GoPro-style wide angle, forward motion, no cuts \|
176	\| Drone flythrough \| Cinematic aerial descent — gimbal-stabilized, sweeping lateral arc, DJI Inspire aesthetic \|
177	\| Architectural flythrough \| Ground-level continuous dolly through connected spaces — one-take, practical lighting \|
178	\| Whip pan \| Very fast horizontal pan with motion blur \|
179	\| Crane shot \| Vertical movement like a crane arm \|
180
181	### Shot Sizes
182	\| Term \| Description \|
183	\|:---\|:---\|
184	\| Extreme close-up \| Eyes, mouth, or small detail only \|
185	\| Close-up \| Face fills frame \|
186	\| Medium close-up \| Head and shoulders \|
187	\| Medium shot \| Waist up \|
188	\| Full shot \| Entire body \|
189	\| Wide / Establishing shot \| Full environment \|
190
191	---
192
193	## 🧠 Prompt Optimization Protocol
194
195	The Agent MUST transform user intent into a technical "Director Brief" before execution.
196
197	1. Technical Grammar: Use camera terms: Dolly In/Out, Crane Shot, Whip Pan, Tracking Shot, Anamorphic Lens, Shallow Depth of Field, High-Speed Dive, Orbital Arc.
198	2. Physics Directives: Use "caustic patterns," "volumetric rays," or "subsurface scattering" instead of "good lighting."
199	3. Timecode Notation: For multi-beat scenes, use `[00:00-00:05s]` format to specify timing.
200	4. Tag References: If files provided, use: "Replicate the camera movement of @video1 while maintaining the visual style of @image1." (lowercase, 1-based index)
201	5. ORDER MATTERS: Tokens at the start define composition; tokens at the end define texture and micro-motion.
202	6. Multi-Image i2v: Provide up to 9 reference images. The model blends aspects (style, identity, environment) across all inputs.
203	7. Audio is mandatory: Seedance 2.0 generates audio natively. Always include an Audio line — music genre/tone, key SFX, ambient texture. Silent direction = random audio.
204	8. Single-beat discipline: Each timed segment = one action. Cramming two narrative beats into 4s degrades physics and motion consistency.
205
206	---
207
208	## 🎭 Capability-Specific Patterns
209
210	### 1. Character Consistency
211	```
212	The man in @Image1 walks tiredly down the hallway, slowing his steps,
213	finally stopping at his front door. Close-up on his face — he takes a
214	deep breath, replaces the weariness with a relaxed expression.
215	Maintain high character consistency, zero facial flicker, persistent clothing details.
216	```
217
218	### 2. Camera Movement Replication
219	```
220	Reference @Image1's male character. He is in @Image2's elevator.
221	Completely reference @Video1's camera movements and facial expressions.
222	Hitchcock zoom during the fear moment, then orbit shots of the interior.
223	Elevator doors open, follow shot walking out.
224	```
225
226	### 3. Video Extension (Forward)
227	```
228	Extend @Video1 by 10 seconds.
229	1–5s: Light and shadow slowly slide across table through venetian blinds.
230	6–10s: A coffee bean drifts down. Camera pushes in toward it until screen goes black.
231	English text gradually appears — "Lucky Coffee", "Breakfast", "AM 7:00-10:00".
232	```
233
234	### 4. Video Extension (Reverse / Prepend)
235	```
236	Extend backward 10s. In warm afternoon light, the camera starts from
237	the corner with awning fluttering in the breeze, slowly tilting down
238	to flowers peeking out at the wall base, building anticipation for the main scene.
239	```
240
241	### 5. Video Editing (Modify Existing)
242	```
243	Subvert @Video1's plot — the character's expression shifts from warmth to
244	cold determination. The action is decisive, without hesitation.
245	Maintain all other visual elements (scene, lighting, timing).
246	```
247
248	### 6. Music Beat-Matching
249	```bash
250	bash scripts/generate-seedance.sh \
251	--mode i2v \
252	--file img1.jpg --file img2.jpg --file img3.jpg \
253	--video-file reference_edit.mp4 \
254	--audio-file track.mp3 \
255	--subject "@Image1 @Image2 @Image3 — match the keyframe positions and rhythm of @Video1 for beat-synced cuts. BGM references @Audio1. More dynamic movement, dreamlike visual style." \
256	--duration 15 --quality high
257	```
258
259	### 7. Dialogue / Voice Acting
260	```
261	In the "Cat & Dog Roast Show" — emotionally expressive comedy segment:
262	Cat host (licking paw, rolling eyes): "Who understands my suffering?"
263	Dog host (head tilted, tail wagging): "You're one to talk? You sleep 18 hours a day..."
264	Sound: lively studio ambience, audience laughter, punchy transitions.
265	```
266
267	### 8. One-Take / Long Take
268	```
269	@Image1 @Image2 @Image3 — one-take tracking shot following a runner
270	from the street up stairs, through a corridor, onto a rooftop,
271	finally overlooking the city. No cuts throughout.
272	```
273
274	### 9. E-commerce / Product Showcase
275	```bash
276	bash scripts/generate-seedance.sh \
277	--mode i2v \
278	--file product.jpg \
279	--subject "Deconstruct the product. Static camera. Hamburger suspended mid-air, rotating slowly. Ingredients separate and reassemble. Cheese continues to melt and drip. Ultimate food aesthetics." \
280	--intent "product" \
281	--aspect "9:16" \
282	--duration 15 --quality high
283	```
284
285	### 10. Science / Educational Visualization
286	```bash
287	bash scripts/generate-seedance.sh \
288	--subject "15-second health educational clip. 0–5s: Transparent blue human upper body, camera pushes into a clear artery, blood flows smoothly. 5–10s: Sugar and fat particles enter bloodstream, lipid deposits form on vessel walls. 10–15s: Vessel narrows, before/after comparison. 4K medical CGI, semi-transparent visualization." \
289	--intent "educational" \
290	--duration 15 --quality high
291	```
292
293	### 11. FPV First-Person Shot
294	```bash
295	bash scripts/generate-seedance.sh \
296	--subject "Immersive first-person POV shot. Camera glides at eye level through a narrow mountain trail,
297	trees rushing past in peripheral blur, rocky terrain below. Slight natural stabilization with wide-angle lens.
298	Continuous forward motion, no cuts throughout. Trail opens into a clearing — mountain peak visible ahead.
299	Sound: wind, footsteps on gravel, distant birds. Natural ambient audio, no music." \
300	--intent "fpv" \
301	--aspect "9:16" --duration 10 --quality high
302	```
303
304	### 12. Cinematic Drone Flythrough
305	```bash
306	bash scripts/generate-seedance.sh \
307	--subject "Cinematic aerial drone shot. Camera starts at 150m altitude above a coastal city at golden hour.
308	Smooth gimbal-stabilized descent along a sweeping lateral arc, dropping toward a rooftop terrace.
309	Long shadows cast across building tops, warm light on ocean surface. High-speed dive closes in
310	to product on the terrace — final frame settles into a medium close-up.
311	Sound: gentle wind, distant city hum, soft cinematic score building to resolve." \
312	--intent "drone" \
313	--aspect "16:9" --duration 10 --tier global --view
314	```
315
316	---
317
318	## 🎨 Prompt Templates
319
320	### Cinematic Film
321	```
322	[SCENE] Rain-soaked cyberpunk alley, neon signs reflected on wet cobblestones.
323	[SUBJECT] A lone figure in a weathered trench coat, face obscured by a wide-brim hat.
324	[ACTION] Walking slowly, each step splashing neon color into the puddles.
325	[CAMERA] Low-angle tracking shot, anamorphic lens, slow dolly in. Rack focus to face.
326	[STYLE] Denis Villeneuve aesthetic, high contrast, desaturated blues and magentas. 24fps.
327	```
328
329	### Product Ad (15s)
330	```
331	Reference @Video1's editing style. Replace @Video1's product with @Image1 as hero.
332	0–3s: Product enters with dynamic rotation, close-up on surface texture and logo.
333	4–8s: Multiple angle transitions — front, side, back — with highlight scanning light.
334	9–12s: Product in lifestyle context showing usage.
335	13–15s: Hero shot with brand tagline, background music builds to resolution.
336	Sound: Reference @Video1's BGM. Add product interaction sound effects.
337	```
338
339	### Short Drama (15s)
340	```
341	Scene (0–5s): Close-up on character's reddened eyes, finger pointing accusingly.
342	Dialogue 1: "What exactly are you trying to take from me?"
343	Scene (6–10s): Other character trembles, holding up evidence, steps forward.
344	Dialogue 2: "I'm not deceiving you! This is what he entrusted to me!"
345	Scene (11–15s): Evidence revealed, first character freezes — anger shifts to shock.
346	Sound: Urgent piano + static interference, sobbing, muffled voice blending in.
347	Duration: Precise 15 seconds, every frame tight, no filler.
348	```
349
350	### Dance / Beat-Sync (13s)
351	```
352	Have the character in @Image1 replicate the dance moves and beat-synced
353	music from @Video1. Generate a 13-second video. Movements should be
354	smooth with no stuttering or freezing.
355	```
356
357	### Scenery Montage (15s)
358	```
359	@Image1 @Image2 @Image3 @Image4 @Image5 @Image6 — landscape scene images.
360	Reference @Video1's visual rhythm, inter-scene transitions, visual style,
361	and music tempo for beat-synced editing.
362	```
363
364	### Advertising / Product Motion
365	```
366	[SCENE] Minimalist white studio, single product on a rotating pedestal.
367	[ACTION] Subtle 360° rotation, product details catching specular highlights.
368	[CAMERA] Tight medium shot, macro lens pass over surface texture, slow orbit.
369	[STYLE] Commercial grade, perfect exposure, zero background distraction.
370	```
371
372	### Action / Physics
373	```
374	[SCENE] Desert canyon at sunrise, sandy terrain, long shadows.
375	[SUBJECT] High-performance sports car accelerating through a turn.
376	[ACTION] Rear wheels spinning with dust plume, chassis flexing under g-force.
377	[CAMERA] Low hero angle dolly tracking alongside, then whip pan to lead car.
378	[STYLE] Hollywood racing film, warm golden grade, motion blur on wheels. 24fps.
379	```
380
381	### Character Consistency (Martial Arts)
382	```
383	[SUBJECT] Same fighter throughout: young woman, white gi, black belt, determined expression.
384	[ACTION] Fluid kata sequence — rising block, stepping side kick, spinning back fist.
385	[CAMERA] Full-body wide shot, then cut to close-up of fist impact in slow motion.
386	[STYLE] Maintain identical lighting, clothing, and facial features in every frame. Zero flicker.
387	```
388
389	---
390
391	## 🎚️ Style & Quality Modifiers
392
393	### Visual Style
394	- `Cinematic quality, film grain, shallow depth of field`
395	- `2.35:1 widescreen, 24fps`
396	- `Ink wash painting style` / `Anime style` / `Photorealistic`
397	- `High saturation neon colors, cool-warm contrast`
398	- `4K medical CGI, semi-transparent visualization`
399
400	### Mood / Atmosphere
401	- `Tense and suspenseful` / `Warm and healing` / `Epic and grand`
402	- `Comedy with exaggerated expressions`
403	- `Documentary tone, restrained narration`
404
405	### Audio Direction
406	- `Background music: grand and majestic`
407	- `Sound effects: footsteps, crowd noise, car sounds`
408	- `Voice tone reference @Video1`
409	- `Beat-synced transitions matching music rhythm`
410
411	---
412
413	## ❌ Common Mistakes to Avoid
414
415	1. Vague references: Don't say "reference @Video1" — specify WHAT to reference (camera? action? effects? rhythm?)
416	2. Conflicting instructions: Don't ask for "static camera" and "orbit shot" in the same segment.
417	3. Overloading: Don't pack too many scenes into 4–5 seconds — keep it physically plausible.
418	4. Missing @ assignments: If you upload 5 images, make sure each one is referenced with a clear purpose.
419	5. Ignoring audio: Sound design dramatically improves output — always include audio direction.
420	6. Forgetting duration: Match prompt complexity to the selected generation length.
421	7. Real faces: Don't upload real human photos — the system will block them.
422	8. Keyword soup: DO NOT use "8k, masterpiece, trending." Use technical descriptions instead.
423	9. Discontinuous action: Avoid "The man runs and then he stops." Use fluid transitional language.
424	10. Missing audio direction: Seedance 2.0 generates audio natively — always specify music tone, SFX, or ambience. Skipping it produces random sound.
425	11. Narrative overload per segment: Each timed segment should contain one action beat. Multiple scene changes in 4s produce degraded physics and motion artifacts.
426	12. FPV without continuous motion: FPV requires a motion-rich environment to work — a static room with FPV intent will not trigger the immersive effect. Pair FPV with corridors, streets, natural terrain, or product flyovers.
427	13. Drone without a destination: Drone shots need a resolve point — specify what the camera descends toward or arrives at. "Drone shot" alone produces aimless floating.
428
429	---
430
431	## 🚀 Protocol: All Modes
432
433	### Mode 1: Text-to-Video (t2v)
434
435	```bash
436	# Chinese tier (default) — epic reveal
437	bash scripts/generate-seedance.sh \
438	--subject "hidden Andes temple, mist through the canopy" \
439	--intent epic --aspect "16:9" --duration 10 --quality high --view
440
441	# Global tier — 21:9 cinematic, 12s
442	bash scripts/generate-seedance.sh \
443	--tier global \
444	--subject "neon cyberpunk alley, rain-slicked streets" \
445	--intent tense --aspect "21:9" --duration 12 --view
446
447	# VIP fast — square social format
448	bash scripts/generate-seedance.sh \
449	--tier vip --fast \
450	--subject "product rotating on a pedestal, specular highlights" \
451	--intent product --aspect "1:1" --duration 6
452	```
453
454	### Mode 2: Image-to-Video (i2v)
455
456	```bash
457	# Chinese tier — animate with video/audio refs
458	bash scripts/generate-seedance.sh --mode i2v \
459	--file character.jpg --video-file ref_motion.mp4 --audio-file bgm.mp3 \
460	--subject "@image1's character walks forward, @video1's camera movement, BGM references @audio1" \
461	--quality high --view
462
463	# Global tier — 1 image = first frame anchor
464	bash scripts/generate-seedance.sh --mode i2v --tier global \
465	--file hero.jpg \
466	--subject "hero strides forward, coat billowing in slow motion" \
467	--duration 8 --view
468
469	# VIP fast — 3 images, omni ref mode (2-9 images switches to omni)
470	bash scripts/generate-seedance.sh --mode i2v --tier vip --fast \
471	--file char.jpg --file env.jpg --file style.jpg \
472	--subject "@image1 character walks through @image2's environment in @image3's style" \
473	--duration 10
474	```
475
476	### Mode 3: Extend Video (Chinese tier)
477
478	```bash
479	# Extend naturally
480	bash scripts/generate-seedance.sh --mode extend \
481	--request-id "abc-123-def-456" --duration 10
482
483	# Extend with directional prompt
484	bash scripts/generate-seedance.sh --mode extend \
485	--request-id "abc-123-def-456" \
486	--subject "camera continues pulling back, revealing the vast city below" \
487	--intent reveal --duration 10 --quality high --view
488	```
489
490	### Mode 4: First & Last Frame (Global/VIP)
491
492	```bash
493	# One image = first frame anchor
494	bash scripts/generate-seedance.sh --mode first-last --tier global \
495	--file opening_scene.jpg \
496	--subject "smooth cinematic push into the scene" --duration 6 --view
497
498	# Two images = interpolate between first and last frame
499	bash scripts/generate-seedance.sh --mode first-last --tier vip --fast \
500	--file start.jpg --file end.jpg \
501	--subject "dramatic reveal transition between the two frames" --duration 8 --view
502	```
503
504	### Mode 5: Omni Reference (omni)
505
506	```bash
507	# Chinese tier — images + video + audio refs
508	bash scripts/generate-seedance.sh --mode omni --tier chinese \
509	--file character.jpg --video-file ref_edit.mp4 --audio-file track.mp3 \
510	--subject "@image1's character performs moves from @video1, BGM references @audio1" \
511	--duration 15 --quality high --view
512
513	# Global tier — images + audio, no video refs
514	bash scripts/generate-seedance.sh --mode omni --tier global \
515	--file portrait.jpg --audio-file bgm.mp3 \
516	--subject "@image1 is the main character. Walking through a neon-lit city at night. BGM references @audio1." \
517	--aspect "16:9" --duration 8 --view
518
519	# VIP tier — with trained omni character
520	bash scripts/generate-seedance.sh --mode omni --tier vip --fast \
521	--subject "@omni-character:char_1775422630065_4vbana walks through a garden at golden hour" \
522	--aspect "16:9" --duration 10 --view
523
524	# With @character ref (from character mode)
525	bash scripts/generate-seedance.sh --mode omni --tier global \
526	--subject "@character:cab9517f-1818-4910-8d66 walks down a rain-soaked alley, cinematic tracking shot" \
527	--duration 8 --view
528	```
529
530	### Mode 6: Train Omni Reference Character (omni-train)
531
532	```bash
533	# Train from a single portrait
534	bash scripts/generate-seedance.sh --mode omni-train \
535	--file portrait.jpg \
536	--character-name "Alex" \
537	--character-desc "A brave explorer with piercing blue eyes"
538
539	# Use in omni prompts after training completes:
540	# @omni-character:<character_id returned>
541	```
542
543	### Mode 7: Character Sheet (character, Chinese tier)
544
545	```bash
546	# Build character from 1–3 reference images
547	bash scripts/generate-seedance.sh --mode character \
548	--file ref1.jpg --file ref2.jpg \
549	--character-name "Hero" \
550	--subject "red leather jacket with black jeans and white sneakers"
551
552	# Use the returned request_id in t2v/i2v/omni:
553	# @character:<request_id>
554	```
555
556	### Mode 8: Video Edit (video-edit, Chinese tier)
557
558	```bash
559	# Replace subject in an existing video
560	bash scripts/generate-seedance.sh --mode video-edit \
561	--video-url "https://example.com/input.mp4" \
562	--file replacement_character.jpg \
563	--subject "Replace the running man with @image1. Preserve exact motion, speed, and camera shake." \
564	--quality high --view
565
566	# Edit with watermark removal in one step
567	bash scripts/generate-seedance.sh --mode video-edit \
568	--video-file source.mp4 \
569	--subject "Subvert the plot — the character's expression shifts from warmth to cold determination." \
570	--remove-watermark --view
571	```
572
573	### Mode 9: Watermark Removal (watermark-remove)
574
575	```bash
576	# Basic watermark removal
577	bash scripts/generate-seedance.sh --mode watermark-remove \
578	--video-url "https://example.com/seedance_output.mp4" --view
579
580	# Pro watermark removal (100MB limit, better quality)
581	bash scripts/generate-seedance.sh --mode watermark-remove \
582	--video-file my_video.mp4 --pro --view
583	```
584
585	### Async Pattern
586
587	```bash
588	# Submit and get request_id immediately
589	RESULT=$(bash scripts/generate-seedance.sh --tier vip --fast --subject "..." --async --json)
590	REQUEST_ID=$(echo "$RESULT" \| jq -r '.request_id')
591
592	# Check status later
593	bash ../../../../core/media/generate-video.sh --result "$REQUEST_ID"
594	```
595
596	---
597
598	## ⚙️ Implementation Details
599
600	### Endpoint Reference
601
602	\| Mode \| Tier \| Endpoint \|
603	\|:---\|:---\|:---\|
604	\| `t2v` \| chinese \| `seedance-v2.0-t2v` \|
605	\| `t2v` \| global \| `seedance-2-text-to-video{-fast}` \|
606	\| `t2v` \| vip \| `seedance-2-vip-text-to-video{-fast}` \|
607	\| `i2v` \| chinese \| `seedance-v2.0-i2v` \|
608	\| `i2v` \| global \| `seedance-2-image-to-video{-fast}` \|
609	\| `i2v` \| vip \| `seedance-2-vip-image-to-video{-fast}` \|
610	\| `extend` \| chinese \| `seedance-v2.0-extend` \|
611	\| `first-last` \| global \| `seedance-2-first-last-frame{-fast}` \|
612	\| `first-last` \| vip \| `seedance-2-vip-first-last-frame{-fast}` \|
613	\| `omni` \| chinese \| `seedance-2.0-omni-reference` \|
614	\| `omni` \| global \| `seedance-2-omni-reference-no-video{-fast}` \|
615	\| `omni` \| vip \| `seedance-2-vip-omni-reference{-fast}` \|
616	\| `omni-train` \| any \| `seedance-2-omni-reference-train` \|
617	\| `character` \| any \| `seedance-2-character` \|
618	\| `video-edit` \| chinese \| `seedance-v2.0-video-edit` \|
619	\| `watermark-remove` \| — \| `seedance-2.0-watermark-remover` / `seedance-2-video-watermark-remover-pro` \|
620
621	### Parameter Differences by Tier
622
623	\| Parameter \| Chinese \| Global \| VIP \|
624	\|:---\|:---\|:---\|:---\|
625	\| `aspect_ratio` \| 16:9, 9:16, 4:3, 3:4 \| + 21:9, 1:1 \| + 21:9, 1:1 \|
626	\| `duration` \| 5 / 10 / 15 (enum) \| 4–15 (any int) \| 4–15 (any int) \|
627	\| `quality` \| basic / high \| — (not supported) \| — (not supported) \|
628	\| `video_files` (omni) \| ✅ up to 3 \| ❌ \| ❌ \|
629	\| `audio_files` (omni) \| ✅ up to 3 \| ✅ up to 3 \| ✅ up to 3 \|
630	\| Fast variant \| ❌ \| ✅ (`--fast`) \| ✅ (`--fast`) \|
631
632	This skill acts as a Cinematographic Wrapper that translates creative intent into high-fidelity technical instructions for the `muapi` core.

Discussion

Alternatives

Also in Training

Act as a Senior Research Paper EvaluatorUse this prompt to critically evaluate academic papers. Ideal for reviewers aiming to assess the quality and validity of scholarly documents across disciplines.Education · CC0-1.0 AstrologerCustom astrologer built in Gemini Gems. Pulls further data from specific websites for a more tailored experience. Can do: astrology & horoscopy, tarot, runes, crystals Possible additions: lenormand, astrology deck, i ching *it's called "Astrologer" because that was the original function, even though it can do more now.Self Improvement · CC0-1.0 Chinese to English Translation AssistantTranslate text from Chinese to English, ensuring accurate and contextually appropriate translations.Education · CC0-1.0 Course Assignment GraderGuide for grading course assignments, including unit test instructions.Education · CC0-1.0