{ "schemaVersion": "1.0", "item": { "slug": "seedance-2-prompt-engineering-skill", "name": "Seedance 2.0 prompt-engineering skill", "source": "tencent", "type": "skill", "category": "AI 智能", "sourceUrl": "https://clawhub.ai/dandysuper/seedance-2-prompt-engineering-skill", "canonicalUrl": "https://clawhub.ai/dandysuper/seedance-2-prompt-engineering-skill", "targetPlatform": "OpenClaw" }, "install": { "downloadMode": "redirect", "downloadUrl": "/downloads/seedance-2-prompt-engineering-skill", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=seedance-2-prompt-engineering-skill", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "installMethod": "Manual import", "extraction": "Extract archive", "prerequisites": [ "OpenClaw" ], "packageFormat": "ZIP package", "includedAssets": [ "_meta.json", "references/modes-and-recipes.md", "references/recipes.md", "references/camera-and-styles.md", "SKILL.sh", "scripts/setup_seedance_prompt_workspace.sh" ], "primaryDoc": "SKILL.md", "quickSetup": [ "Download the package from Yavira.", "Extract the archive and review SKILL.md first.", "Import or place the package into your OpenClaw setup." ], "agentAssist": { "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.", "steps": [ "Download the package from Yavira.", "Extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the extracted folder." ], "prompts": [ { "label": "New install", "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete." }, { "label": "Upgrade existing", "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run." } ] }, "sourceHealth": { "source": "tencent", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-04-30T16:55:25.780Z", "expiresAt": "2026-05-07T16:55:25.780Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentDisposition": "attachment; filename=\"network-1.0.0.zip\"", "redirectLocation": null, "bodySnippet": null }, "scope": "source", "summary": "Source download looks usable.", "detail": "Yavira can redirect you to the upstream package for this source.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/seedance-2-prompt-engineering-skill" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] }, "downloadPageUrl": "https://openagent3.xyz/downloads/seedance-2-prompt-engineering-skill", "agentPageUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill/agent", "manifestUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill/agent.json", "briefUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill/agent.md" }, "agentAssist": { "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.", "steps": [ "Download the package from Yavira.", "Extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the extracted folder." ], "prompts": [ { "label": "New install", "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete." }, { "label": "Upgrade existing", "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run." } ] }, "documentation": { "source": "clawhub", "primaryDoc": "SKILL.md", "sections": [ { "title": "Purpose", "body": "Create high-control English prompts for Seedance 2.0 and Seedance 2.0 Fast using multimodal references (image/video/audio/text).\n\nThis skill is for:\n\nPrompt design from rough idea to production-ready prompt\nMode choice: Text-only vs First/Last Frame vs All-Reference\n@asset mapping (what each image/video/audio controls)\n4-15s duration planning and timeline beats\nMulti-segment stitching for videos >15s\nVideo extension / continuation prompts\nCharacter replacement and directed editing prompts\nCamera-language replication from reference videos\nScenario-specific strategies (product ads, short drama, fantasy, music video, etc.)" }, { "title": "Core Rules", "body": "Always declare mode first.\nAlways include an explicit Assets Mapping section.\nUse timecoded beats with one major action per segment.\nKeep prompts concise and controllable (avoid vague poetic-only wording).\nAdd negative constraints when user needs clean output.\nBe specific and visual — \"a woman in a red trench coat walks through rain-soaked neon streets\" >> \"a woman walking\".\nSeparate dialogue/sound from visuals — write dialogue with character name + emotion tag, then sound effects as a distinct layer.\nMatch reference image style to video theme — e.g., ink-wash style images for historical themes, neon renders for cyberpunk." }, { "title": "Platform Limits (Seedance 2.0)", "body": "Mixed inputs total (image+video+audio): max 12 files\nImages: jpeg/png/webp/bmp/tiff/gif, max 9, each < 30MB\nVideos: mp4/mov, max 3, total duration 2-15s, total < 50MB\nAudio: mp3/wav, max 3, total <= 15s, total < 15MB\nGeneration duration: 4-15s\nRealistic human face references may be blocked by platform compliance" }, { "title": "Output Format (use by default)", "body": "Mode\nAssets Mapping\nFinal Prompt\nNegative Constraints\nGeneration Settings\n\nExample skeleton:\n\nMode: All-Reference\nAssets Mapping:\n- @image1: first frame / identity anchor\n- @video1: camera language + motion rhythm\n- @audio1: optional soundtrack pacing\n\nFinal Prompt:\n[ratio], [duration], [style].\n0-3s: [action + camera].\n3-7s: [action + transition].\n7-10s: [reveal/climax + end frame].\nPreserve identity and scene continuity. Use physically plausible motion and coherent lighting.\n\nNegative Constraints:\nno watermark, no logo, no subtitles, no on-screen text.\n\nGeneration Settings:\nDuration: 10s\nAspect Ratio: 9:16" }, { "title": "IP / Copyright Avoidance (Moderation-Safe Prompting)", "body": "Seedance 2.0 has platform-side content moderation. Prompts referencing recognizable franchises, characters, or brand aesthetics will be rejected even if no name is used. Follow these rules:" }, { "title": "Core Principles", "body": "Never use franchise names, character names, or brand terms — not even as \"style of\" references.\nInvent fully original names for characters and creatures. Use descriptive nicknames (e.g., \"Alloy Sentinel\", \"Storm-Rabbit\").\nDescribe aesthetics generically — replace recognizable signature features with original alternatives:\n\n❌ \"arc reactor\" → ✅ \"hex-light energy core\"\n❌ \"yellow lightning mouse\" → ✅ \"tiny storm-rabbit with glowing cyan antlers\"\n❌ \"red-gold armored suit\" → ✅ \"custom exo-suit with smooth ceramic panels\"\n\n\nAdd explicit negative constraints listing every franchise name, character name, and brand term that could be inferred.\nUse family-friendly / PG-13 tone markers — they help pass moderation." }, { "title": "Progressive Fallback Strategy", "body": "If a prompt is rejected, escalate distance from the source IP:\n\nLevel 1: Replace all names with original nicknames, keep general aesthetic.\nLevel 2: Replace signature visual features (colors, silhouette, iconic props) with fully original designs.\nLevel 3: Change character type entirely (e.g., humanoid hero → autonomous mech + drone; creature battle → abstract elemental spirits)." }, { "title": "Toy / Figure Animation", "body": "When animating toy or doll references from images:\n\nStrip all brand indicators from the prompt.\nUse \"original vinyl-style toy figure\" or \"collectible art figure\" instead of any brand name.\nBind @image1 to proportions, colors, outfit shape only — never preserve logos or trademarks." }, { "title": "A) Extend Video", "body": "Explicitly write: Extend @video1 by Xs.\nUse generation duration equal to the newly added segment, not the full final length." }, { "title": "B) Replace Character", "body": "Bind base motion/camera to @video1, bind replacement identity to @image1, and request strict choreography/timing preservation." }, { "title": "C) Beat Sync", "body": "Use @video/@audio rhythm references and lock beats by time range." }, { "title": "D) Text-Only Generation", "body": "Use when no reference assets are provided. Prompt must carry all visual direction: style, color palette, character descriptions, camera, and timeline beats. Especially useful for original creature/character concepts and IP-safe scenes." }, { "title": "E) Multi-Segment Stitching (Videos > 15s)", "body": "Seedance 2.0 max generation is 15s per segment. For longer videos, split into chained segments:\n\nSegment 1: Generate normally (up to 15s). End on a clean handoff frame (stable pose, clear composition).\nSegment 2+: Upload previous segment as @video1, write Extend @video1 by Xs. Include a continuity note describing exactly what the last frame looks like.\nRepeat until target duration is reached.\n\nAlways include:\n\nTotal duration and segment count at the top.\nHandoff description at the end of each segment (what the last frame shows).\nExplicit continuity instructions: preserve identity, outfit, lighting, camera direction." }, { "title": "F) Short Drama with Dialogue", "body": "For scripted scenes with character speech:\n\nWrite visual action and dialogue as separate layers per time segment.\nTag dialogue: Dialogue (CharacterName, emotion): \"line\"\nTag sound: Sound: [description]\nKeep dialogue short — one line per 3-5s segment works best." }, { "title": "G) Product Showcase / E-Commerce Ad", "body": "For product demos and ads:\n\nBind product image to @image1 as identity anchor.\nUse techniques: 360° rotation, 3D exploded view, reassembly animation, hero lighting.\nKeep background clean (studio, gradient, or contextual lifestyle).\nSpecify material rendering: glass reflections, metallic sheen, matte texture, etc." }, { "title": "H) One-Take Long Shot (Multi-Image Waypoints)", "body": "For continuous tracking shots without cuts:\n\nAssign each @image to a scene waypoint (location, character, or prop encountered along the path).\nWrite the prompt as a continuous camera movement visiting each waypoint in order.\nExplicitly state: no cuts, single continuous shot or one-take.\nUse @image1 as first frame, subsequent images as reference for environments/characters encountered." }, { "title": "Scenario-Specific Strategies", "body": "ScenarioKey TechniquesTypical ModeE-commerce / Product Ad360° spin, 3D exploded view, hero lighting, clean studio BGAll-ReferenceShort Drama / DialogueDialogue tags with emotion, sound FX layer, actor blockingAll-Reference or First FrameFantasy / Xianxia AnimationSpell FX particles, martial arts choreography, energy aurasText-only or All-ReferenceScience / Education4K CGI, transparent anatomy, labeled zoom sequencesText-onlyMusic Video / Beat SyncBeat-locked cuts, widescreen 16:9, multi-image montageAll-Reference with @audioOne-Take Tracking ShotMulti-image waypoints, continuous camera, no cutsAll-ReferenceIP-Safe Original CharactersInvented names, unique features, explicit negative constraintsText-only" }, { "title": "Files in this skill", "body": "SKILL.md — main skill behavior\nSKILL.sh — quick local test helper\nscripts/setup_seedance_prompt_workspace.sh — scaffold helper files\nreferences/recipes.md — ready-to-use prompt recipes\nreferences/modes-and-recipes.md — mode and control notes\nreferences/camera-and-styles.md — camera language and visual styles vocabulary" } ], "body": "Seedance 2.0 JiMeng Skill (OpenClaw / ClawHub)\nPurpose\n\nCreate high-control English prompts for Seedance 2.0 and Seedance 2.0 Fast using multimodal references (image/video/audio/text).\n\nThis skill is for:\n\nPrompt design from rough idea to production-ready prompt\nMode choice: Text-only vs First/Last Frame vs All-Reference\n@asset mapping (what each image/video/audio controls)\n4-15s duration planning and timeline beats\nMulti-segment stitching for videos >15s\nVideo extension / continuation prompts\nCharacter replacement and directed editing prompts\nCamera-language replication from reference videos\nScenario-specific strategies (product ads, short drama, fantasy, music video, etc.)\nCore Rules\nAlways declare mode first.\nAlways include an explicit Assets Mapping section.\nUse timecoded beats with one major action per segment.\nKeep prompts concise and controllable (avoid vague poetic-only wording).\nAdd negative constraints when user needs clean output.\nBe specific and visual — \"a woman in a red trench coat walks through rain-soaked neon streets\" >> \"a woman walking\".\nSeparate dialogue/sound from visuals — write dialogue with character name + emotion tag, then sound effects as a distinct layer.\nMatch reference image style to video theme — e.g., ink-wash style images for historical themes, neon renders for cyberpunk.\nPlatform Limits (Seedance 2.0)\nMixed inputs total (image+video+audio): max 12 files\nImages: jpeg/png/webp/bmp/tiff/gif, max 9, each < 30MB\nVideos: mp4/mov, max 3, total duration 2-15s, total < 50MB\nAudio: mp3/wav, max 3, total <= 15s, total < 15MB\nGeneration duration: 4-15s\nRealistic human face references may be blocked by platform compliance\nOutput Format (use by default)\nMode\nAssets Mapping\nFinal Prompt\nNegative Constraints\nGeneration Settings\n\nExample skeleton:\n\nMode: All-Reference\nAssets Mapping:\n- @image1: first frame / identity anchor\n- @video1: camera language + motion rhythm\n- @audio1: optional soundtrack pacing\n\nFinal Prompt:\n[ratio], [duration], [style].\n0-3s: [action + camera].\n3-7s: [action + transition].\n7-10s: [reveal/climax + end frame].\nPreserve identity and scene continuity. Use physically plausible motion and coherent lighting.\n\nNegative Constraints:\nno watermark, no logo, no subtitles, no on-screen text.\n\nGeneration Settings:\nDuration: 10s\nAspect Ratio: 9:16\n\nIP / Copyright Avoidance (Moderation-Safe Prompting)\n\nSeedance 2.0 has platform-side content moderation. Prompts referencing recognizable franchises, characters, or brand aesthetics will be rejected even if no name is used. Follow these rules:\n\nCore Principles\nNever use franchise names, character names, or brand terms — not even as \"style of\" references.\nInvent fully original names for characters and creatures. Use descriptive nicknames (e.g., \"Alloy Sentinel\", \"Storm-Rabbit\").\nDescribe aesthetics generically — replace recognizable signature features with original alternatives:\n❌ \"arc reactor\" → ✅ \"hex-light energy core\"\n❌ \"yellow lightning mouse\" → ✅ \"tiny storm-rabbit with glowing cyan antlers\"\n❌ \"red-gold armored suit\" → ✅ \"custom exo-suit with smooth ceramic panels\"\nAdd explicit negative constraints listing every franchise name, character name, and brand term that could be inferred.\nUse family-friendly / PG-13 tone markers — they help pass moderation.\nProgressive Fallback Strategy\n\nIf a prompt is rejected, escalate distance from the source IP:\n\nLevel 1: Replace all names with original nicknames, keep general aesthetic.\nLevel 2: Replace signature visual features (colors, silhouette, iconic props) with fully original designs.\nLevel 3: Change character type entirely (e.g., humanoid hero → autonomous mech + drone; creature battle → abstract elemental spirits).\nToy / Figure Animation\n\nWhen animating toy or doll references from images:\n\nStrip all brand indicators from the prompt.\nUse \"original vinyl-style toy figure\" or \"collectible art figure\" instead of any brand name.\nBind @image1 to proportions, colors, outfit shape only — never preserve logos or trademarks.\nSpecial Cases\nA) Extend Video\n\nExplicitly write: Extend @video1 by Xs. Use generation duration equal to the newly added segment, not the full final length.\n\nB) Replace Character\n\nBind base motion/camera to @video1, bind replacement identity to @image1, and request strict choreography/timing preservation.\n\nC) Beat Sync\n\nUse @video/@audio rhythm references and lock beats by time range.\n\nD) Text-Only Generation\n\nUse when no reference assets are provided. Prompt must carry all visual direction: style, color palette, character descriptions, camera, and timeline beats. Especially useful for original creature/character concepts and IP-safe scenes.\n\nE) Multi-Segment Stitching (Videos > 15s)\n\nSeedance 2.0 max generation is 15s per segment. For longer videos, split into chained segments:\n\nSegment 1: Generate normally (up to 15s). End on a clean handoff frame (stable pose, clear composition).\nSegment 2+: Upload previous segment as @video1, write Extend @video1 by Xs. Include a continuity note describing exactly what the last frame looks like.\nRepeat until target duration is reached.\n\nAlways include:\n\nTotal duration and segment count at the top.\nHandoff description at the end of each segment (what the last frame shows).\nExplicit continuity instructions: preserve identity, outfit, lighting, camera direction.\nF) Short Drama with Dialogue\n\nFor scripted scenes with character speech:\n\nWrite visual action and dialogue as separate layers per time segment.\nTag dialogue: Dialogue (CharacterName, emotion): \"line\"\nTag sound: Sound: [description]\nKeep dialogue short — one line per 3-5s segment works best.\nG) Product Showcase / E-Commerce Ad\n\nFor product demos and ads:\n\nBind product image to @image1 as identity anchor.\nUse techniques: 360° rotation, 3D exploded view, reassembly animation, hero lighting.\nKeep background clean (studio, gradient, or contextual lifestyle).\nSpecify material rendering: glass reflections, metallic sheen, matte texture, etc.\nH) One-Take Long Shot (Multi-Image Waypoints)\n\nFor continuous tracking shots without cuts:\n\nAssign each @image to a scene waypoint (location, character, or prop encountered along the path).\nWrite the prompt as a continuous camera movement visiting each waypoint in order.\nExplicitly state: no cuts, single continuous shot or one-take.\nUse @image1 as first frame, subsequent images as reference for environments/characters encountered.\nScenario-Specific Strategies\nScenario\tKey Techniques\tTypical Mode\nE-commerce / Product Ad\t360° spin, 3D exploded view, hero lighting, clean studio BG\tAll-Reference\nShort Drama / Dialogue\tDialogue tags with emotion, sound FX layer, actor blocking\tAll-Reference or First Frame\nFantasy / Xianxia Animation\tSpell FX particles, martial arts choreography, energy auras\tText-only or All-Reference\nScience / Education\t4K CGI, transparent anatomy, labeled zoom sequences\tText-only\nMusic Video / Beat Sync\tBeat-locked cuts, widescreen 16:9, multi-image montage\tAll-Reference with @audio\nOne-Take Tracking Shot\tMulti-image waypoints, continuous camera, no cuts\tAll-Reference\nIP-Safe Original Characters\tInvented names, unique features, explicit negative constraints\tText-only\nFiles in this skill\nSKILL.md — main skill behavior\nSKILL.sh — quick local test helper\nscripts/setup_seedance_prompt_workspace.sh — scaffold helper files\nreferences/recipes.md — ready-to-use prompt recipes\nreferences/modes-and-recipes.md — mode and control notes\nreferences/camera-and-styles.md — camera language and visual styles vocabulary" }, "trust": { "sourceLabel": "tencent", "provenanceUrl": "https://clawhub.ai/dandysuper/seedance-2-prompt-engineering-skill", "publisherUrl": "https://clawhub.ai/dandysuper/seedance-2-prompt-engineering-skill", "owner": "dandysuper", "version": "2.0.0", "license": null, "verificationStatus": "Indexed source record" }, "links": { "detailUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill", "downloadUrl": "https://openagent3.xyz/downloads/seedance-2-prompt-engineering-skill", "agentUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill/agent", "manifestUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill/agent.json", "briefUrl": "https://openagent3.xyz/skills/seedance-2-prompt-engineering-skill/agent.md" } }