Skip to content
Back to search

seedance

SkillWorks with Claude CodeGitHub

Generate cinematic motion graphics videos using ByteDance Seedance 2.0 via Fal AI. Covers image-to-video, reference-to-video, text-to-video, liquid glass prompting, App Store screenshot scraping, and the full pipeline from screenshot to finished promo clip. Triggers on seedance, cdance, motion graphics, liquid glass, app promo, video promo.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/robonuggets--seedance-skill/badge.svg)](/listings/robonuggets--seedance-skill)

HTML

<a href="/listings/robonuggets--seedance-skill"><img src="/api/listings/robonuggets--seedance-skill/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

1. Install the skill

Copy the `SKILL.md` file to your Claude Code project's `.claude/skills/seedance/` directory.

2. Add your Fal AI key

echo "FAL_KEY=your-fal-ai-key" >> .env

Get a key at [fal.ai](https://fal.ai).

3. Generate

Tell Claude Code:

Generate a liquid glass motion graphics promo for https://linear.app

Claude will screenshot the site, upload the images, generate the video, and download it.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 668 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. ---
  2. name: seedance
  3. description: Generate cinematic motion graphics videos using ByteDance Seedance 2.0 via Fal AI. Covers image-to-video, reference-to-video, text-to-video, liquid glass prompting, App Store screenshot scraping, and the full pipeline from screenshot to finished promo clip. Triggers on seedance, cdance, motion graphics, liquid glass, app promo, video promo.
  4. ---
  5. # Seedance — Seedance 2.0 Motion Graphics Skill
  6. Generate cinematic motion graphics videos using ByteDance's Seedance 2.0 model via Fal AI. Turn app screenshots into glossy liquid glass promo videos controlled from Claude Code.
  7. ## What This Skill Does
  8. - Generate **image-to-video** animations from a single image (with optional end frame control)
  9. - Generate **reference-to-video** motion graphics from multiple app screenshots
  10. - Generate **text-to-video** motion graphics from pure prompts
  11. - Scrape **App Store screenshots** automatically for any iOS app
  12. - Upload reference images to get **public URLs** for the API
  13. - Apply proven **liquid glass prompting patterns** that produce Apple-keynote-quality output
  14. ## Setup
  15. ### 1. Get a Fal AI API Key
  16. Sign up at [fal.ai](https://fal.ai) and create an API key from your dashboard.
  17. ### 2. Store the Key
  18. Add to your `.env` file:
  19. ```
  20. FAL_KEY=your-fal-ai-key-here
  21. ```
  22. ### 3. File Upload (Included)
  23. Seedance needs **publicly accessible URLs** for reference images. Fal AI includes its own CDN upload — no extra accounts or keys needed. Your `FAL_KEY` handles both generation and file upload.
  24. > **Want a full Generations tab with gallery, styles panel, and multi-model support (Kling, Veo, GPT Image, Nano Banana)?** Check out the Rubric dashboard at [robonuggets.com](https://robonuggets.com).
  25. ---
  26. ## Endpoints
  27. Seedance 2.0 has 5 endpoints on Fal AI. **Default to Pro** — it's only ~25% more expensive but noticeably better quality.
  28. | Type | Tier | Endpoint | $/sec (720p) |
  29. |------|------|----------|-------------|
  30. | **Image-to-video** | **Pro** | `fal.run/bytedance/seedance-2.0/image-to-video` | $0.30 |
  31. | Reference-to-video | **Pro (default)** | `fal.run/bytedance/seedance-2.0/reference-to-video` | $0.30 |
  32. | Reference-to-video | Fast (testing) | `fal.run/bytedance/seedance-2.0/fast/reference-to-video` | $0.24 |
  33. | Text-to-video | **Pro (default)** | `fal.run/bytedance/seedance-2.0/text-to-video` | $0.30 |
  34. | Text-to-video | Fast (testing) | `fal.run/bytedance/seedance-2.0/fast/text-to-video` | $0.24 |
  35. ### When to Use Which
  36. - **Image-to-video** — you have a single image and want to animate it with motion. Supports start + end frame control for precise transitions. Best for: animating a hero image, creating A→B transitions between two states, lip-sync animations.
  37. - **Reference-to-video** — you have multiple screenshots, logos, or visual assets to blend into a motion graphics piece. This is the primary mode for promo videos.
  38. - **Text-to-video** — no visual references available (e.g. GitHub repos, docs pages, abstract concepts). Pure prompt-driven.
  39. ---
  40. ## Parameters
  41. | Param | Required | Options | Default |
  42. |-------|----------|---------|---------|
  43. | `prompt` | Yes | string | — |
  44. | `image_urls` | No | array of public URLs, up to 9 | — |
  45. | `video_urls` | No | array of public URLs, up to 3 | — |
  46. | `audio_urls` | No | array of public URLs, up to 3 | — |
  47. | `resolution` | No | `"480p"`, `"720p"` | `"720p"` |
  48. | `duration` | No | `"auto"`, `"4"` – `"15"` | `"auto"` |
  49. | `aspect_ratio` | No | `"auto"`, `"21:9"`, `"16:9"`, `"4:3"`, `"1:1"`, `"3:4"`, `"9:16"` | `"auto"` |
  50. | `generate_audio` | No | boolean | `true` |
  51. | `seed` | No | integer | random |
  52. - **Total files across all modalities must not exceed 12**
  53. - Reference images in prompts: `@Image1`, `@Image2`, ... `@Image9`
  54. - Reference videos: `@Video1`, `@Video2`, `@Video3`
  55. - Reference audio: `@Audio1`, `@Audio2`, `@Audio3`
  56. - Max resolution is **720p** — there is no 1080p option
  57. ### Cost Reference (Pro, 720p)
  58. | Duration | Cost |
  59. |----------|------|
  60. | 4s | $1.21 |
  61. | 5s | $1.51 |
  62. | 8s | $2.42 |
  63. | 10s | $3.03 |
  64. | 12s | $3.63 |
  65. | 15s | $4.54 |
  66. Fast tier is ~20% cheaper. Use for quick tests before committing to Pro.
  67. ---
  68. ## Sweet Spot (Tested)
  69. These are the proven ratios from extensive testing:
  70. | Refs | Duration | Use Case | Cost (Pro 720p) |
  71. |------|----------|----------|-----------------|
  72. | 1 ref | 4s | Quick test, single screenshot | $1.21 |
  73. | 3 refs | 4–6s | Fast promo test | $1.21–$1.81 |
  74. | **5 refs** | **10s** | **Production promo (best results)** | **$3.03** |
  75. | 3 refs | 4s 480p | Cheapest possible test | ~$0.81 |
  76. **The golden rule: 1 reference image per 2 seconds of video.**
  77. - Fewer refs = model focuses better on each one
  78. - More refs (6-7+) = model blends too loosely, loses clarity
  79. - Always describe each image's role in the prompt
  80. ---
  81. ## Prompting Guide
  82. ### Rule 1: Logo References First (CRITICAL)
  83. **NEVER describe logos or brand elements in the prompt text.** Seedance mangles text rendering — wrong shapes, wrong colors, hallucinated details.
  84. Before writing ANY prompt:
  85. 1. Ask for or find the app/brand logo image
  86. 2. Save it locally
  87. 3. Upload it to get a public URL
  88. 4. Pass it as one of the `@Image` references
  89. If you don't have the logo: **STOP and ask.** Never proceed without it.
  90. **Recommended approach:** Generate the motion graphics without text/logos in the video at all, then overlay the logo in post-production (Premiere, After Effects, etc). This produces the cleanest results.
  91. ### Rule 2: The Liquid Glass Prompt Template
  92. This is the proven prompt structure that produces Apple-keynote-quality motion graphics:
  93. ```
  94. Design a motion-graphics style ad using glossy liquid glass design
  95. language. Pure black background. @Image1 is [describe what it shows].
  96. @Image2 is [describe]. @Image3 is [describe]. The [element] from
  97. Image 1 becomes translucent glass [describe transformation].
  98. [Element] from Image 2 [describe]. [Continue for each image].
  99. Multiple camera angles: close-up on glass reflections, pull back
  100. to reveal the full scene. No text, no logos, no words. Style: liquid
  101. glass morphism, Apple Vision Pro aesthetic, premium 3D depth,
  102. self-luminous forms on absolute black, [accent color] accent lighting.
  103. ```
  104. **Key rules:**
  105. - Describe each `@Image` by number and what it contains
  106. - Describe how each element transforms into glass
  107. - End with a style line including the brand's accent color
  108. - `"Multiple camera angles"` keeps it dynamic
  109. - `"No text, no logos, no words"` prevents garbled text
  110. - Keep under ~200 words — the model ignores long prompts
  111. ### Rule 3: Device Framing
  112. For product shots, specify the device in the prompt:
  113. **Floating desktop screens (best results):**
  114. ```
  115. Show the desktop screens floating in 3D space on a pure black
  116. background, tilted at slight angles like a MacBook product shot.
  117. The UI elements on screen become translucent glass with reflections
  118. and refractions.
  119. ```
  120. **iPad:**
  121. ```
  122. An iPad Pro floating in empty black space, tilted at a cinematic
  123. angle like an Apple product shot. The iPad has visible black bezels
  124. and a physical frame — only the screen content has the glass effect.
  125. ```
  126. **MacBook:**
  127. ```
  128. A MacBook Pro floating in empty black space, open at a cinematic
  129. angle. The MacBook screen displays the dashboard. Light catches
  130. the aluminium edges.
  131. ```
  132. **Floating in black void** produces better results than placing devices in environments (offices, desks). Keep it pure black.
  133. ### Rule 4: Audio
  134. - `generate_audio: true` adds ambient SFX and music
  135. - May trigger **content policy violations** on abstract visuals — if it fails, retry with `generate_audio: false`
  136. - Audio costs the same whether enabled or not
  137. - When it works, the generated audio adds real production value
  138. ### Rule 5: Negative Guidance
  139. Add constraints to prevent common issues:
  140. - `"No text, no logos, no words"` — prevents garbled text
  141. - `"No cuts"` — keeps it as one continuous shot
  142. - `"No camera shake"` — keeps it stable
  143. - `"Pure visual motion only"` — reinforces no text
  144. ---
  145. ## The Full Pipeline
  146. ### Step 1: Get Screenshots
  147. **Option A — Screenshot a website:**
  148. ```python
  149. from playwright.sync_api import sync_playwright
  150. with sync_playwright() as p:
  151. browser = p.chromium.launch(headless=True)
  152. page = browser.new_page(viewport={"width": 1920, "height": 1080})
  153. page.goto("https://example.com", wait_until="networkidle", timeout=30000)
  154. page.wait_for_timeout(2000)
  155. page.screenshot(path="screenshot.png")
  156. browser.close()
  157. ```
  158. **Option B — Scrape App Store screenshots:**
  159. ```python
  160. import json, urllib.request
  161. # Step 1: Find the app ID
  162. search_url = "https://itunes.apple.com/search?term=APP_NAME&entity=software&limit=3"
  163. data = json.loads(urllib.request.urlopen(search_url).read())
  164. app_url = data["results"][0]["trackViewUrl"]
  165. # Step 2: Scrape screenshot URLs from the page using Playwright
  166. # Look for <source> elements with srcset containing "mzstatic.com"
  167. # The URLs use format like /460x996bb.webp — swap to /1290x2796bb.jpg for full res
  168. ```
  169. **Option C — Manual screenshots** — take them yourself, save locally.
  170. ### Step 2: Upload for Public URL
  171. Seedance needs publicly accessible URLs for reference images. Use Fal AI's built-in CDN upload (default) — no extra accounts needed.
  172. **Fal AI CDN (default — uses your existing FAL_KEY):**
  173. ```bash
  174. # Step 1: Initiate upload
  175. INIT=$(curl -s -X POST "https://rest.alpha.fal.ai/storage/upload/initiate" \
  176. -H "Authorization: Key $FAL_KEY" \
  177. -H "Content-Type: application/json" \
  178. -d '{"file_name":"my-screenshot.png","content_type":"image/png"}')
  179. # Extract URLs from response
  180. FILE_URL=$(echo "$INIT" | python -c "import sys,json; print(json.load(sys.stdin)['file_url'])")
  181. UPLOAD_URL=$(echo "$INIT" | python -c "import sys,json; print(json.load(sys.stdin)['upload_url'])")
  182. # Step 2: Upload the file
  183. curl -s -X PUT "$UPLOAD_URL" \
  184. -H "Content-Type: image/png" \
  185. --data-binary @/path/to/screenshot.png
  186. ```
  187. Use `FILE_URL` as the public URL in your Seedance API calls. Hosted on `v3b.fal.media` with long-lived caching. Included with your Fal AI account — no extra cost.
  188. **Alternative: Kie AI (if you have a key):**
  189. ```bash
  190. curl -s -X POST "https://kieai.redpandaai.co/api/file-stream-upload" \
  191. -H "Authorization: Bearer $KIE_API_KEY" \
  192. -F "file=@/path/to/screenshot.png" \
  193. -F "uploadPath=generations" \
  194. -F "fileName=my-screenshot.png"
  195. ```
  196. Returns `downloadUrl`. Files expire after 3 days.
  197. **Alternative: Litterbox (no account needed):**
  198. ```bash
  199. curl -s -F "reqtype=fileupload" -F "time=1h" \
  200. -F "fileToUpload=@/path/to/screenshot.png" \
  201. https://litterbox.catbox.moe/resources/internals/api.php
  202. ```
  203. Returns a direct URL. Expires after 1 hour.
  204. ### Step 3: Generate the Video
  205. **Always use a temp JSON file** to avoid shell escaping issues:
  206. ```bash
  207. cat > /tmp/seedance_body.json << 'ENDJSON'
  208. {
  209. "prompt": "Design a motion-graphics style ad using glossy liquid glass design language. Pure black background. @Image1 is [describe]. @Image2 is [describe]. @Image3 is [describe]. The elements become translucent glass with reflections and refractions. Multiple camera angles. No text, no logos, no words. Style: liquid glass morphism, Apple Vision Pro aesthetic, premium 3D depth, self-luminous forms on absolute black.",
  210. "image_urls": [
  211. "https://your-upload-url/screenshot1.png",
  212. "https://your-upload-url/screenshot2.png",
  213. "https://your-upload-url/screenshot3.png"
  214. ],
  215. "resolution": "720p",
  216. "duration": "4",
  217. "aspect_ratio": "16:9",
  218. "generate_audio": false
  219. }
  220. ENDJSON
  221. curl -s -X POST "https://fal.run/bytedance/seedance-2.0/reference-to-video" \
  222. -H "Authorization: Key $FAL_KEY" \
  223. -H "Content-Type: application/json" \
  224. -d @/tmp/seedance_body.json
  225. ```
  226. ### Step 4: Download the Result
  227. The response contains a video URL:
  228. ```json
  229. {
  230. "video": { "url": "https://v3b.fal.media/files/.../video.mp4" },
  231. "seed": 42
  232. }
  233. ```
  234. **Download immediately** — URLs may expire.
  235. ```bash
  236. curl -sL -o "output_video.mp4" "THE_VIDEO_URL"
  237. ```
  238. ### Step 5: Post-Production (Recommended)
  239. Overlay the logo and any text in your video editor. The motion graphics are the visual base — branding goes on top in post.
  240. ---
  241. ## Text-to-Video (No References)
  242. For apps/projects where you don't have screenshots (GitHub repos, docs pages, abstract concepts), use text-to-video with the timestamped approach:
  243. ```
  244. Ultra-sleek motion design sequence. Pure black background throughout.
  245. Glossy liquid glass design language.
  246. 0-1s [FORM]: A translucent glass [object] materializes at center,
  247. light refracting through it.
  248. 1-2s [CONNECT]: Glass connections form between elements, each a thin
  249. glass tube filling with luminous light.
  250. 2-3s [REVEAL]: The structure rotates revealing depth and glass
  251. reflections.
  252. 3-4s [GLOW]: A pulse of warm light passes through the entire glass
  253. structure.
  254. Style: liquid glass morphism, premium 3D depth, self-luminous forms
  255. on absolute black.
  256. ```
  257. **Endpoint:** `https://fal.run/bytedance/seedance-2.0/text-to-video` (Pro)
  258. Same parameters minus `image_urls`, `video_urls`, `audio_urls`.
  259. ---
  260. ## Image-to-Video (Single Image Animation)
  261. Animate a single still image into cinematic video. Unlike reference-to-video (which blends multiple images into motion graphics), image-to-video takes **one image** and brings it to life with motion described in the prompt.
  262. **Endpoint:** `https://fal.run/bytedance/seedance-2.0/image-to-video`
  263. ### Parameters
  264. | Param | Required | Options | Default |
  265. |-------|----------|---------|---------|
  266. | `prompt` | Yes | string — describe the desired motion/action | — |
  267. | `image_url` | Yes | single public URL (JPEG, PNG, WebP, max 30MB) | — |
  268. | `end_image_url` | No | public URL for the last frame (enables A→B transitions) | — |
  269. | `resolution` | No | `"480p"`, `"720p"` | `"720p"` |
  270. | `duration` | No | `"auto"`, `"4"` – `"15"` | `"auto"` |
  271. | `aspect_ratio` | No | `"auto"`, `"21:9"`, `"16:9"`, `"4:3"`, `"1:1"`, `"3:4"`, `"9:16"` | `"auto"` |
  272. | `generate_audio` | No | boolean — includes SFX, ambient, and lip-sync | `true` |
  273. | `seed` | No | integer | random |
  274. ### Key Differences from Reference-to-Video
  275. - **Single `image_url`** (string) instead of `image_urls` (array)
  276. - **`end_image_url`** — optional end frame for controlled transitions (reference-to-video doesn't have this)
  277. - **Lip-sync support** — audio generation includes lip-synced speech when faces are present
  278. - No `@Image1`/`@Image2` notation — just one image as the starting frame
  279. ### When to Use Image-to-Video vs Reference-to-Video
  280. | Scenario | Use |
  281. |----------|-----|
  282. | Animate a single screenshot/photo | **Image-to-video** |
  283. | Transition between two states (before/after, light/dark mode) | **Image-to-video** with `end_image_url` |
  284. | Lip-sync a character or avatar | **Image-to-video** |
  285. | Blend 3-5 screenshots into a promo reel | Reference-to-video |
  286. | Glass morphism motion graphics from multiple UI views | Reference-to-video |
  287. ### Usage Example
  288. ```bash
  289. cat > /tmp/seedance_i2v.json << 'ENDJSON'
  290. {
  291. "prompt": "The dashboard interface comes alive — UI elements gently float and shift, glass reflections ripple across panels, accent lighting pulses through the layout. Smooth cinematic camera drift. No text, no logos.",
  292. "image_url": "https://your-upload-url/screenshot.png",
  293. "resolution": "720p",
  294. "duration": "4",
  295. "aspect_ratio": "16:9",
  296. "generate_audio": false
  297. }
  298. ENDJSON
  299. curl -s -X POST "https://fal.run/bytedance/seedance-2.0/image-to-video" \
  300. -H "Authorization: Key $FAL_KEY" \
  301. -H "Content-Type: application/json" \
  302. -d @/tmp/seedance_i2v.json
  303. ```
  304. ### Start + End Frame Example (A→B Transition)
  305. Animate from one state to another — e.g. light mode to dark mode, empty dashboard to populated:
  306. ```bash
  307. cat > /tmp/seedance_i2v_transition.json << 'ENDJSON'
  308. {
  309. "prompt": "The interface smoothly transitions — colors shift, panels rearrange, elements morph fluidly from the starting layout to the final layout. Cinematic, smooth, continuous motion.",
  310. "image_url": "https://your-upload-url/state_before.png",
  311. "end_image_url": "https://your-upload-url/state_after.png",
  312. "resolution": "720p",
  313. "duration": "6",
  314. "aspect_ratio": "16:9",
  315. "generate_audio": false
  316. }
  317. ENDJSON
  318. curl -s -X POST "https://fal.run/bytedance/seedance-2.0/image-to-video" \
  319. -H "Authorization: Key $FAL_KEY" \
  320. -H "Content-Type: application/json" \
  321. -d @/tmp/seedance_i2v_transition.json
  322. ```
  323. ### Prompting Tips for Image-to-Video
  324. - **Describe the motion**, not the image — the model already sees the image. Focus on what should move, how, and where.
  325. - **"Cinematic camera drift"** works well for subtle product shots
  326. - **"Elements gently float and shift"** for UI animation
  327. - For lip-sync: describe the speech/emotion in the prompt, enable `generate_audio: true`
  328. - Same negative guidance applies: `"No text, no logos, no words"` to prevent garbled text
  329. - The `end_image_url` feature is powerful for before/after reveals, mode switches, or showing a workflow progression
  330. ---
  331. ## Prompt Patterns Library
  332. ### Floating Desktop Screens (Best for SaaS/Apps)
  333. ```
  334. Design a motion-graphics style ad based on these images, using
  335. multiple camera angles — close-ups that highlight the glass
  336. reflections, various tap and click interaction animations, and a
  337. focus on this glossy glass UI design language. @Image1 is [describe].
  338. @Image2 is [describe]. @Image3 is [describe]. Show the desktop
  339. screens floating in 3D space on a pure black background, tilted at
  340. slight angles like a MacBook product shot. The UI elements on screen
  341. become translucent glass with reflections and refractions. No text,
  342. no logos, no words. Style: liquid glass morphism, Apple product ad,
  343. premium desktop app showcase, self-luminous forms on absolute black,
  344. [color] accent lighting.
  345. ```
  346. ### Glass UI Ad (Based on LEPADPHONE Style)
  347. ```
  348. Design a motion-graphics style ad based on this image, using multiple
  349. camera angles — close-ups that highlight the glass reflections,
  350. various tap/click interaction animations, and a focus on this glossy
  351. "glass UI" design language. @Image1 is [describe the app/interface].
  352. The interface elements become translucent glass panels with
  353. refractions. No text, no logos. Style: liquid glass morphism, premium
  354. depth, self-luminous forms on absolute black.
  355. ```
  356. ### Floating Desktop Screens — Rubric (Tested)
  357. Used for the Rubric dashboard promo. 3 refs (Flows + Skill Tree + Agent Icons), Pro 720p 4s, audio on.
  358. ```
  359. Design a motion-graphics style ad based on these images, using
  360. multiple camera angles — close-ups that highlight the glass
  361. reflections, various tap and click interaction animations, and a
  362. focus on this glossy glass UI design language. @Image1 is a desktop
  363. application dashboard showing agent workflow connections on a
  364. hexagonal grid. @Image2 is a desktop application showing a network
  365. graph visualization with connected nodes. @Image3 is a desktop
  366. application showing a grid of colorful pixel art icons. Show the
  367. desktop screens floating in 3D space on a pure black background,
  368. tilted at slight angles like a MacBook product shot. The UI elements
  369. on screen become translucent glass with reflections and refractions.
  370. No text, no logos, no words. Style: liquid glass morphism, Apple
  371. product ad, premium desktop app showcase, self-luminous forms on
  372. absolute black, orange accent lighting.
  373. ```
  374. ### iPad Product Reveal — Rubric (Tested)
  375. Real device with glass screen content. 2 refs (hero + Flows), Pro 720p 4s, audio on. Key detail: explicitly call out "real solid device" so the iPad frame stays solid while only the screen gets glass effect.
  376. ```
  377. Design a motion-graphics style ad using glossy liquid glass design
  378. language. Pure black background, nothing else. A realistic iPad Pro
  379. with visible black bezels and physical frame floating in empty black
  380. space, tilted at a cinematic angle like an Apple product shot. The
  381. iPad is a real solid device with clear borders and edges, not made
  382. of glass. Only the screen content has the glass effect. The iPad
  383. screen displays @Image1 and @Image2 — a dark dashboard with orange
  384. branding, hexagonal grid, and agent flow connections. The UI elements
  385. on the screen are translucent glass with reflections and refractions.
  386. Light catches the physical edges of the iPad frame. The device
  387. slowly rotates. No text, no logos, no words. Style: Apple keynote
  388. product reveal, real device with glass UI on screen, premium 3D
  389. depth, floating in absolute black void, orange accent lighting.
  390. ```
  391. ### Abstract Glass Network — Rubric (Tested)
  392. No device frame — pure abstract glass. 3 refs (Flows + Icons + Skill Tree), Fast 720p 4s, no audio.
  393. ```
  394. Design a motion-graphics style ad using glossy liquid glass design
  395. language. Pure black background. @Image1 is a dark dashboard with
  396. hexagonal grid and agent connections. @Image2 is a grid of colorful
  397. pixel art bot icons. @Image3 is a network graph of agents with
  398. radiating connections. The hexagonal grid from Image 1 becomes
  399. translucent glass honeycomb, each cell refracting light. Colorful
  400. bot icons from Image 2 appear as glass tiles catching light. The
  401. network from Image 3 pulses with glass fiber-optic connections,
  402. light flowing between nodes. Multiple camera angles: close-up on
  403. glass reflections, pull back to reveal the full network. Style:
  404. liquid glass morphism, Apple Vision Pro aesthetic, premium 3D depth,
  405. self-luminous forms on absolute black, orange accent lighting.
  406. ```
  407. ### Abstract Motion Identity (Based on @oggii_0 Style)
  408. ```
  409. Ultra-sleek motion design sequence. Pure black background throughout.
  410. White-on-black minimal aesthetic. No cuts — one single continuous
  411. unbroken visual evolution. Locked-off static center frame, no camera
  412. movement. All motion occurs within the subject. Self-luminous white
  413. forms on absolute black, no grain, no shadows.
  414. 0-2s [ORIGIN]: [Starting visual]
  415. 2-4s [TRANSFORM]: [Transformation]
  416. 4-6s [REVEAL]: [Reveal]
  417. 6-8s [CLOSE]: [Closing visual]
  418. Style: [brand] motion identity, mathematical precision.
  419. Mood: calm, inevitable, intelligent, premium.
  420. ```
  421. ---
  422. ## Batch Generation Workflow
  423. When generating promos for multiple projects at once, follow this process:
  424. ### Step 1: Assess Each Project
  425. For each app/project, determine:
  426. - **Do screenshots exist?** Website or App Store → reference-to-video. GitHub repo or docs-only → text-to-video.
  427. - **What's the visual anchor?** Dark UI = strong liquid glass match. Colorful icons = glass tiles. Geometric layout = glass honeycomb.
  428. - **What aspect ratio?** Phone app → 9:16. Desktop/SaaS → 16:9. Abstract → 16:9.
  429. ### Step 2: Test Batch (Cheap)
  430. Run all projects at **Fast tier, 480p, 4 seconds** first. This keeps each test under ~$0.81. For a 5-project batch, total test cost is ~$4-5.
  431. ### Step 3: Review & Promote Winners
  432. Watch all test outputs. The ones that work well, re-generate at **Pro tier, 720p, longer duration** (8-10s). This is where you spend the real budget.
  433. ### Step 4: Post-Production
  434. Overlay logos and text in your video editor. The Seedance output is the visual base — branding goes on top in post.
  435. ### Example: 5-Project Test Batch
  436. | # | Project | Type | Rationale |
  437. |---|---------|------|-----------|
  438. | 1 | GitHub repo (no UI) | text-to-video | No screenshots, abstract concept |
  439. | 2 | Dark landing page | ref-to-video | Dark hero screenshot = strong glass match |
  440. | 3 | Docs site | text-to-video | No strong visual to reference |
  441. | 4 | GitHub repo (abstract) | text-to-video | Geometric concept works well prompt-only |
  442. | 5 | iOS app | ref-to-video | App Store screenshots scraped + logo ref |
  443. All Fast, 480p, 4s (except iOS app at 6s for more screen time). Total: ~$5.50.
  444. ---
  445. ## Cost Estimation
  446. Before generating, estimate your cost:
  447. ### Quick Formula
  448. ```
  449. Cost = duration_seconds × $0.30 (Pro) or $0.24 (Fast) at 720p
  450. ```
  451. 480p is roughly 30% cheaper than 720p.
  452. ### Estimation Table
  453. | What you're doing | Recommended settings | Est. cost |
  454. |-------------------|---------------------|-----------|
  455. | Quick concept test | Fast, 480p, 4s | ~$0.81 |
  456. | Single screenshot test | Pro, 720p, 4s | ~$1.21 |
  457. | Short promo clip | Pro, 720p, 6s | ~$1.81 |
  458. | Production promo | Pro, 720p, 10s | ~$3.03 |
  459. | Long showcase | Pro, 720p, 15s | ~$4.54 |
  460. | 5-project test batch | Fast, 480p, 4-6s each | ~$4-6 |
  461. | 3-video production set | Pro, 720p, 4s each | ~$3.63 |
  462. ### Before You Generate Checklist
  463. 1. **Pick your tier** — Fast for testing, Pro for production
  464. 2. **Pick your resolution** — 480p for tests, 720p for keepers
  465. 3. **Pick your duration** — 4s minimum, 10s sweet spot, 15s max
  466. 4. **Count your refs** — 1 image per 2 seconds of video (max 5)
  467. 5. **Audio on or off?** — Same cost either way, but abstract visuals may trigger content policy. Default off for tests.
  468. 6. **Multiply it out** — duration × rate = cost per video
  469. ---
  470. ## Gotchas & Known Issues
  471. 1. **Text rendering is bad** — never ask Seedance to render text. Overlay in post.
  472. 2. **Audio content policy** — abstract visuals may trigger false positives. Default to `generate_audio: false`, enable when needed.
  473. 3. **People's faces in screenshots** — may trigger "likenesses of real people" content policy. Crop faces out of screenshots.
  474. 4. **Shell escaping breaks JSON** — always use temp JSON files with `cat > /tmp/file.json`, never inline JSON in curl commands.
  475. 5. **Kie AI upload URLs expire after 3 days** — re-upload if generating later.
  476. 6. **Max resolution is 720p** — no 1080p. Upscale in post if needed.
  477. 7. **Pro vs Fast** — Pro is only ~25% more expensive but noticeably better. Default to Pro.
  478. 8. **7+ reference images** — quality degrades. Stick to 5 or fewer.
  479. ---
  480. ## Auth Header
  481. Fal AI uses `Authorization: Key` (not Bearer):
  482. ```
  483. Authorization: Key YOUR_FAL_KEY
  484. ```
  485. ---
  486. ## Community Prompts
  487. For inspiration beyond liquid glass, check the community collection:
  488. **[awesome-seedance-2-prompts](https://github.com/YouMind-OpenLab/awesome-seedance-2-prompts)** — 1,719 curated Seedance 2.0 prompts covering cinematic, anime, commercial, and experimental styles. Includes video examples and a searchable web gallery at [youmind.com](https://youmind.com/en-US/seedance-2-0-prompts).
  489. ---
  490. ## Credits
  491. Built by [RoboNuggets](https://robonuggets.com). Tested extensively with Rubric dashboard screenshots, App Store scraping, and liquid glass motion graphics.
  492. Seedance 2.0 is by ByteDance. Fal AI is the API provider.
README.md · README · 86 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Seedance Skill
  2. > Turn any app into a cinematic motion graphics promo video using AI.
  3. Seedance Skill is a Claude Code skill that generates glossy liquid glass motion graphics videos using ByteDance's Seedance 2.0 model via Fal AI. Give it screenshots of any app, and it produces Apple-keynote-quality promo clips.
  4. ## Examples
  5. - Screenshot a SaaS dashboard -> 10-second liquid glass promo video
  6. - Scrape App Store screenshots -> motion graphics ad for any iOS app
  7. - Text prompt -> abstract motion identity sequence
  8. ## Quick Start
  9. ### 1. Install the skill
  10. Copy the `SKILL.md` file to your Claude Code project's `.claude/skills/seedance/` directory.
  11. ### 2. Add your Fal AI key
  12. ```bash
  13. echo "FAL_KEY=your-fal-ai-key" >> .env
  14. ```
  15. Get a key at [fal.ai](https://fal.ai).
  16. ### 3. Generate
  17. Tell Claude Code:
  18. ```
  19. Generate a liquid glass motion graphics promo for https://linear.app
  20. ```
  21. Claude will screenshot the site, upload the images, generate the video, and download it.
  22. ## What You Get
  23. - **4-second test clips** for ~$1.21 (Pro, 720p)
  24. - **10-second production promos** for ~$3.03 (Pro, 720p)
  25. - Glossy liquid glass aesthetic with translucent panels, light refractions, and depth
  26. - Multiple camera angles, floating devices, pure black backgrounds
  27. ## Prompting Patterns
  28. The skill includes battle-tested prompt templates:
  29. - **Liquid Glass** — translucent UI panels with reflections (the default)
  30. - **Floating Desktop** — MacBook/iPad product shots in black void
  31. - **Glass UI Ad** — close-ups on glass interactions
  32. - **Abstract Motion Identity** — timestamped sequences with geometric transformations
  33. ## Sweet Spot
  34. | Refs | Duration | Use Case | Cost (Pro 720p) |
  35. |------|----------|----------|-----------------|
  36. | 3 refs | 4s | Quick test | $1.21 |
  37. | **5 refs** | **10s** | **Production promo** | **$3.03** |
  38. | 1 ref | 4s 480p | Cheapest test | ~$0.81 |
  39. **Golden rule: 1 reference image per 2 seconds of video.**
  40. ## Key Rules
  41. 1. **Never describe logos in prompts** — pass the actual logo image as a reference, or overlay in post
  42. 2. **Default to Pro tier** — only 25% more than Fast but noticeably better
  43. 3. **Use temp JSON files** for API calls — shell escaping breaks curl
  44. 4. **No text in video** — overlay text/logos in post-production for clean results
  45. 5. **Max 5 reference images** — more than that and quality degrades
  46. ## Pricing
  47. Seedance 2.0 on Fal AI:
  48. | Tier | Cost/sec (720p) |
  49. |------|-----------------|
  50. | Pro (default) | $0.30/sec |
  51. | Fast (testing) | $0.24/sec |
  52. ## Built By
  53. [RoboNuggets](https://robonuggets.com) — AI agent tools and community.
  54. ## License
  55. MIT

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.