baseball-trend
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
Chinesebaseball-trend
baseball-trend
15-second ESPN-style broadcast cutaway of a user, sitting behind home plate at a fake Yankees vs Red Sox ALCS Game 3 game at Fenway Park, with two announcers naming them on air.
Fixed-recipe skill — the prompts below are calibrated. Substitute the username and keep the marked anchors intact.
生成一段15秒ESPN风格的转播镜头,画面中用户坐在芬威公园内假洋基队对阵红袜队ALCS第三场的本垒后方座位上,两位解说员会在直播中提及用户姓名。
这是一个固定流程的技能——以下提示词已校准。替换用户名时请保留标记锚点不变。
Stage 0 — Intake
阶段0 — 信息收集
If invoked with empty args and no usable prior context, print this menu and stop:
Who should appear in the fake MLB broadcast cutaway? Required:
- Name — exactly as it should appear in the chyron and announcer dialogue
- Reference photo — one front-facing or 3/4 portrait, local path or HTTPS URL
If only one field is missing, ask only for that field. Otherwise ask the two questions below one at a time.
1. Username (required) — used both in the broadcast chyron and in the announcers' commentary, e.g. . Save as . This replaces every literal in the prompts below.
"Jane Doe"state.username${username}2. Reference image (required) — one front-facing 3/4 portrait, good lighting, one face. Resolve to a CDN URL and save as :
state.reference_image_url-
URL → use as-is.
https://… -
Local path → upload the file with, then use the returned public URL.
upload_asset -
Claude Desktop, photo pasted inline → inline pastes don't reach MCP tools yet (Anthropic limitation). Reply with:Heads up — pasted images don't reach MCP tools on Claude Desktop yet. Two options:
- Paste a URL if it's already hosted somewhere — fastest.
- Attach the image file so I can upload it before generation.
When a local file arrives: convert it to a public URL withand useupload_asset.public_url
After both answers are in, echo one short confirmation ("Generating behind-home-plate cutaway for {username}…") and start the pipeline. No further yes/no gates after this point — the pipeline runs end-to-end.
如果调用时参数为空且无可用上下文,显示以下菜单并停止:
谁将出现在假MLB转播镜头中? 必填项:
- 姓名 — 需与字幕条和解说台词中显示的完全一致
- 参考照片 — 一张正面或3/4角度的肖像照,提供本地路径或HTTPS链接
如果仅缺少一个字段,仅询问该字段。否则依次询问以下两个问题。
1. 用户名 (必填) — 用于转播字幕条和解说台词,例如 。保存为 。需替换下方提示词中所有 的字面内容。
"Jane Doe"state.username${username}2. 参考图片 (必填) — 一张正面或3/4角度的肖像照,光线良好,仅含一张人脸。转换为CDN链接并保存为 :
state.reference_image_url-
链接 → 直接使用。
https://… -
本地路径 → 使用上传文件,然后使用返回的公共链接。
upload_asset -
Claude桌面端粘贴的图片 → 当前内联粘贴的图片无法传输到MCP工具(Anthropic限制)。回复:注意:Claude桌面端粘贴的图片目前无法传输到MCP工具。有两个选项:
- 如果图片已托管在某处,请粘贴链接——这是最快的方式。
- 请附加图片文件,以便我在生成前上传。
当收到本地文件时:使用转换为公共链接,并使用upload_asset。public_url
收集完两个信息后,输出简短确认语(“正在为**{username}**生成本垒后转播镜头…”)并启动流程。此后无需再确认——流程将全程自动运行。
Pipeline
流程
Two Pika MCP calls, sequential. Engines are locked: for the still, for the video.
gpt-image-2kling-v3-omni两次Pika MCP调用,按顺序执行。引擎固定:静态图使用 ,视频使用 。
gpt-image-2kling-v3-omniStep 1 — Broadcast still (generate_image
)
generate_image步骤1 — 转播静态图 (generate_image
)
generate_imageThe chyron + scorebug get baked into the still at frame 0 (load-bearing — when Kling is asked to "pop in" the chyron mid-clip it appears at second 4–5 with a visible flash and breaks the trend; baking it into the first frame makes Kling treat it as pixel-locked burned-in UI).
Call with:
mcp__pika__generate_image- :
providergpt-image-2 - :
reference_images[state.reference_image_url] - :
aspect_ratio16:9 - :
quality(default for speed;mediumis now exposed but ~2 min/call — use only when fidelity matters)high - :
output_formatpng - (verbatim,
promptsubstituted):${username}
A screenshot from a live MLB game TV broadcast on ESPN. The camera cuts to the audience — our reference image person, sitting smiling in premium field-level seats behind home plate at Fenway Park, smiling naturally and unaware they're on camera. Hardlock: Do not alter their facial structure and maintain their likeness. The subject must match the reference person.
The image looks exactly like a real TV screenshot — broadcast color grading, slight compression artifacts, interlacing grain, telephoto broadcast camera feel. It's the New York Yankees vs Boston Red Sox, MLB American League Championship Series (ALCS), Game 3, Boston home stadium (Fenway Park). Yankees lead 2-0 in the ALCS so far.
CRITICAL — broadcast graphics that MUST be visible in this image:
1. A real ESPN-style bottom scorebug for MLB, showing Yankees vs Red Sox with team logos, inning, outs, balls/strikes count, and score (with a small runners-on-base diamond), looking like a real live broadcast scorebug.
2. Directly above the scorebug, a clean broadcast-style lower-third name graphic / chyron that reads exactly: "${username}" — set in a classic ESPN sans-serif, in the network's color treatment. The chyron sits in the lower-left area, above the scorebug, like a real broadcast identifier for the on-camera guest.
3. The ESPN network logo watermark in a corner.
All three graphics must look like real burned-in broadcast UI — not Photoshop overlays. 16:9 aspect ratio.Save the returned URL as .
state.broadcast_still_urlAgent-side self-check before Step 2: the chyron must spell the username correctly and the scorebug must look like real broadcast UI. If either looks wrong, re-roll Step 1 (everything downstream pixel-locks to this frame). This is the agent's own check — do not ask the user.
字幕条和计分牌需嵌入第0帧的静态图中(至关重要——如果要求Kling在视频中途添加字幕条,会在第4-5秒出现明显闪烁,破坏特效;将其嵌入第一帧可让Kling将其视为像素锁定的内置UI)。
调用 ,参数如下:
mcp__pika__generate_image- :
providergpt-image-2 - :
reference_images[state.reference_image_url] - :
aspect_ratio16:9 - :
quality(默认选项以提升速度;medium选项已开放,但每次调用约需2分钟——仅在对保真度有要求时使用)high - :
output_formatpng - (原样使用,替换
prompt):${username}
A screenshot from a live MLB game TV broadcast on ESPN. The camera cuts to the audience — our reference image person, sitting smiling in premium field-level seats behind home plate at Fenway Park, smiling naturally and unaware they're on camera. Hardlock: Do not alter their facial structure and maintain their likeness. The subject must match the reference person.
The image looks exactly like a real TV screenshot — broadcast color grading, slight compression artifacts, interlacing grain, telephoto broadcast camera feel. It's the New York Yankees vs Boston Red Sox, MLB American League Championship Series (ALCS), Game 3, Boston home stadium (Fenway Park). Yankees lead 2-0 in the ALCS so far.
CRITICAL — broadcast graphics that MUST be visible in this image:
1. A real ESPN-style bottom scorebug for MLB, showing Yankees vs Red Sox with team logos, inning, outs, balls/strikes count, and score (with a small runners-on-base diamond), looking like a real live broadcast scorebug.
2. Directly above the scorebug, a clean broadcast-style lower-third name graphic / chyron that reads exactly: "${username}" — set in a classic ESPN sans-serif, in the network's color treatment. The chyron sits in the lower-left area, above the scorebug, like a real broadcast identifier for the on-camera guest.
3. The ESPN network logo watermark in a corner.
All three graphics must look like real burned-in broadcast UI — not Photoshop overlays. 16:9 aspect ratio.将返回的链接保存为 。
state.broadcast_still_url步骤2前的Agent自检:字幕条中的用户名拼写必须正确,计分牌外观需与真实转播UI一致。如果其中任意一项有误,重新执行步骤1(后续所有环节均基于此帧像素锁定)。此检查由Agent自动完成——无需询问用户。
Step 2 — 15s broadcast video (generate_reference_video
)
generate_reference_video步骤2 — 15秒转播视频 (generate_reference_video
)
generate_reference_videoimage_types: ["first_frame"]state.broadcast_still_urlCall with:
mcp__pika__generate_reference_video- :
providerkling - :
kling_modelkling-v3-omni - :
duration15 - :
aspect_ratio16:9 - :
quality_modepro - :
reference_images[state.broadcast_still_url] - :
image_types["first_frame"] - :
soundtrue - :
prompt_adherence(load-bearing — without it the scorebug animates and identity drifts late in the clip)strict - (verbatim, load-bearing — without these entries Kling occasionally morphs the scorebug or fades the chyron):
negative_prompt
scene cuts, camera angle changes, scorebug animation, chyron pop-in, chyron fade-in, chyron text changes, graphics animating, exaggerated acting, direct address to camera, blurry face, identity drift, distorted anatomy- (verbatim,
promptsubstituted everywhere; pre-trimmed to fit Kling's 2500-char cap; chyron-on-frame-0 lock at top):${username}
First frame is the provided reference image. The ESPN scorebug AND the "${username}" lower-third chyron are ALREADY on screen at frame 0 — keep them visible, unchanged, pixel-locked across all 15 seconds. Do NOT animate them, do NOT change their text.
Realistic live MLB broadcast shot of the subject sitting in premium field-level seats behind home plate at Yankees vs Red Sox ALCS Game 3 in Boston, Fenway Park. The shot feels like a real TV cutaway when the broadcast camera finds a notable guest in the crowd between innings.
The subject is seated in his field-level seat, smiling naturally and not over-performing. Not locked into eye contact with the lens. Occasionally glances toward the field, then toward camera, then back to the field — like a real in-game crowd reaction. One continuous take. No cuts. No angle changes.
Action timeline:
0-4s: smiling casually in his seat as the camera lands on him; looks around naturally, not paying attention to camera.
4-7s: relaxed natural wave toward the camera (crowd cheers when he waves the first time); glances up at the Jumbotron above him then back to camera.
7-11s: cheers briefly with visible excitement, reacting to the playoff atmosphere; turns to his friend on the left, exchanges words, laughs (we don't hear him speak).
11-15s: claps naturally while smiling.
Keep all movement subtle, believable, human. No exaggerated acting. No direct talking to camera.
Broadcast styling: real live sports broadcast look, telephoto broadcast camera feel, natural ballpark lighting, slight broadcast compression, slight interlacing / TV grain, authentic crowd movement in the background, realistic field-level framing. Subject remains seated behind home plate the full shot.
Audio: Natural live sports-broadcast commentary from two male announcers talking about him being at the game tonight. Casual, warm, authentic — like real MLB commentators noticing a known guest. Sample lines:
"${username} is here tonight at Fenway, taking in this massive playoff matchup."
"You can see he's enjoying himself here behind home plate for Game 3."
"Great atmosphere in the building, and ${username} getting a lot of love from the crowd."
Constraints: Preserve identity strongly. Keep him seated behind home plate throughout. No constant eye contact with camera. No talking to camera. No exaggerated gestures. No scene cuts. Scorebug + chyron do not change at any point. Genuine MLB TV broadcast crowd cutaway feel.Save the returned video URL as . If generation completes asynchronously, follow the MCP tool's returned status handle until the video reaches a terminal state.
state.broadcast_video_urlimage_types: ["first_frame"]state.broadcast_still_url调用 ,参数如下:
mcp__pika__generate_reference_video- :
providerkling - :
kling_modelkling-v3-omni - :
duration15 - :
aspect_ratio16:9 - :
quality_modepro - :
reference_images[state.broadcast_still_url] - :
image_types["first_frame"] - :
soundtrue - :
prompt_adherence(至关重要——如果不设置此参数,计分牌会出现动画,且后期镜头中人物身份会偏移)strict - (原样使用,至关重要——如果缺少这些条目,Kling偶尔会让计分牌变形或字幕条淡化):
negative_prompt
scene cuts, camera angle changes, scorebug animation, chyron pop-in, chyron fade-in, chyron text changes, graphics animating, exaggerated acting, direct address to camera, blurry face, identity drift, distorted anatomy- (原样使用,替换所有
prompt;已预先修剪以适配Kling的2500字符限制;顶部锁定第0帧字幕条):${username}
First frame is the provided reference image. The ESPN scorebug AND the "${username}" lower-third chyron are ALREADY on screen at frame 0 — keep them visible, unchanged, pixel-locked across all 15 seconds. Do NOT animate them, do NOT change their text.
Realistic live MLB broadcast shot of the subject sitting in premium field-level seats behind home plate at Yankees vs Red Sox ALCS Game 3 in Boston, Fenway Park. The shot feels like a real TV cutaway when the broadcast camera finds a notable guest in the crowd between innings.
The subject is seated in his field-level seat, smiling naturally and not over-performing. Not locked into eye contact with the lens. Occasionally glances toward the field, then toward camera, then back to the field — like a real in-game crowd reaction. One continuous take. No cuts. No angle changes.
Action timeline:
0-4s: smiling casually in his seat as the camera lands on him; looks around naturally, not paying attention to camera.
4-7s: relaxed natural wave toward the camera (crowd cheers when he waves the first time); glances up at the Jumbotron above him then back to camera.
7-11s: cheers briefly with visible excitement, reacting to the playoff atmosphere; turns to his friend on the left, exchanges words, laughs (we don't hear him speak).
11-15s: claps naturally while smiling.
Keep all movement subtle, believable, human. No exaggerated acting. No direct talking to camera.
Broadcast styling: real live sports broadcast look, telephoto broadcast camera feel, natural ballpark lighting, slight broadcast compression, slight interlacing / TV grain, authentic crowd movement in the background, realistic field-level framing. Subject remains seated behind home plate the full shot.
Audio: Natural live sports-broadcast commentary from two male announcers talking about him being at the game tonight. Casual, warm, authentic — like real MLB commentators noticing a known guest. Sample lines:
"${username} is here tonight at Fenway, taking in this massive playoff matchup."
"You can see he's enjoying himself here behind home plate for Game 3."
"Great atmosphere in the building, and ${username} getting a lot of love from the crowd."
Constraints: Preserve identity strongly. Keep him seated behind home plate throughout. No constant eye contact with camera. No talking to camera. No exaggerated gestures. No scene cuts. Scorebug + chyron do not change at any point. Genuine MLB TV broadcast crowd cutaway feel.将返回的视频链接保存为 。如果生成为异步执行,需跟踪MCP工具返回的状态句柄,直到视频进入终态。
state.broadcast_video_urlStep 3 — Deliver
步骤3 — 交付
Return both Pika CDN URLs: the still image URL and the final video URL. If the host client requires local media markers, create the local preview outside this skill after confirming both CDN URLs are reachable.
One-line summary: "Behind-home-plate cutaway for {username} — 15s, 16:9, 1080p, Kling v3-omni, native two-announcer commentary."
返回两个Pika CDN链接:静态图链接和最终视频链接。如果宿主客户端需要本地媒体标记,确认两个CDN链接可访问后,在此技能外创建本地预览。
一句话总结:"Behind-home-plate cutaway for {username} — 15s, 16:9, 1080p, Kling v3-omni, native two-announcer commentary."
Load-bearing phrases (don't strip these)
关键短语(请勿删除)
These are empirical behavior dependencies, not writing style — removing them breaks the recipe:
- In the still prompt: +
Hardlock: Do not alter their facial structure and maintain their likeness(without these, identity drifts on the first frame, and everything downstream inherits the drift).The subject must match the reference person - In the video prompt: +
Preserve identity strongly(without these, Kling re-animates the chyron mid-clip).The ESPN scorebug AND the "${username}" lower-third chyron are ALREADY on screen at frame 0 — keep them visible, unchanged, pixel-locked - The full negative_prompt list — every entry there came from a specific failure mode in prior runs.
- and
prompt_adherence: "strict"— see inline notes above.image_types: ["first_frame"]
这些是经验证的行为依赖项,而非写作风格——删除它们会导致流程失效:
- 在静态图提示词中:+
Hardlock: Do not alter their facial structure and maintain their likeness(缺少这些,第一帧人物身份会偏移,后续所有环节都会继承该偏移)。The subject must match the reference person - 在视频提示词中:+
Preserve identity strongly(缺少这些,Kling会在视频中途重新制作字幕条动画)。The ESPN scorebug AND the "${username}" lower-third chyron are ALREADY on screen at frame 0 — keep them visible, unchanged, pixel-locked - 完整的negative_prompt列表——每个条目都来自之前运行中出现的特定故障模式。
- 和
prompt_adherence: "strict"——参见上方内联注释。image_types: ["first_frame"]
Engine choice: Kling-only (with one caveat)
引擎选择:仅使用Kling(有一个注意事项)
Seedance has a two-stage 422 gate (validated 2026-05-12 across 4 runs on the NBA sibling skill):
partner_validation_failed- Input-side (): rejects if the reference contains a recognizable real person.
body.image_urls - Output-side (): rejects AFTER generation if the produced clip contains recognizable-looking faces — and every broadcast cutaway has a crowd full of faces.
body.generated_video
The output-side gate is unavoidable for this trend regardless of subject, so Seedance is functionally unusable here. Kling is the engine that works for ordinary user photos.
Kling caveat — recognizable celebrities are blocked too. Kling has its own content-moderation gate that fires on celebrity references (validated 2026-05-13: a Michael Jordan reference + "Ke Wang" chyron returned at submit-time). This is correct behavior — the trend illusion only works with a non-public-figure reference where the chyron name + face are coherent. If a user supplies a celebrity photo, surface the gate to them and ask for a non-celebrity reference instead.
task_status: failed, task_status_msg: "Failure to pass the risk control system"Kling trade-offs: 2500-char cap (recipe above is pre-trimmed), no param (re-rolls are non-reproducible — to re-roll just call again).
promptseedSeedance存在两阶段 422限制(2026年5月12日通过NBA同类技能的4次运行验证):
partner_validation_failed- 输入端 ():如果参考图片包含可识别的真实人物,会被拒绝。
body.image_urls - 输出端 ():生成后如果视频包含可识别的人脸,会被拒绝——而所有转播镜头的观众席都充满人脸。
body.generated_video
无论主体是谁,输出端限制都无法避免,因此Seedance在此场景下无法使用。Kling是唯一适用于普通用户照片的引擎。
Kling注意事项——可识别的名人也会被屏蔽。Kling有自己的内容审核限制,会触发名人参考的拦截(2026年5月13日验证:使用迈克尔·乔丹的参考图片+“Ke Wang”字幕条,返回 )。这是正确的行为——该特效只有在使用非公众人物的参考图片,且字幕条姓名与面部一致时才能生效。如果用户提供名人照片,需告知该限制并要求更换为非名人参考图片。
task_status: failed, task_status_msg: "Failure to pass the risk control system"Kling的权衡:有2500字符限制(上方流程已预先修剪),无参数(重新生成不可重复——只需再次调用即可重新生成)。
promptseedRuntime expectations
运行时间预期
Typical run time is 4-7 minutes:
| Step | Wall clock | Notes |
|---|---|---|
| Reference upload | 5-30s | Skip when the user supplies HTTPS |
| Broadcast still | 60-120s | Re-roll before video if the chyron or scorebug is wrong |
| Kling video | 3-5 min | One 15s pro render with native commentary |
| Delivery check | <30s | Verify final URL and obvious identity/chyron continuity |
典型运行时间为4-7分钟:
| 步骤 | 耗时 | 说明 |
|---|---|---|
| 参考图片上传 | 5-30秒 | 用户提供HTTPS链接时可跳过 |
| 转播静态图 | 60-120秒 | 如果字幕条或计分牌有误,需在生成视频前重新执行此步骤 |
| Kling视频生成 | 3-5分钟 | 一次15秒专业渲染,带原生解说 |
| 交付检查 | <30秒 | 验证最终链接及明显的身份/字幕条连续性 |
Failure modes
故障模式
| Symptom | Cause | Fix |
|---|---|---|
| Chyron pops in mid-clip (~4–5s flash) | Chyron not baked into the still | Re-run Step 1; verify chyron is visible in |
| Scorebug animates / morphs mid-clip | | Restore strict adherence and the full negative_prompt |
| Identity drift late in the clip (face changes after ~10s) | Reference image too small / Kling losing the face | Re-run Step 2; if drift persists, re-run Step 1 with a tighter face crop on the still (more facial pixels = stronger lock) |
| Username mispronounced by announcers | Native audio is one take | Re-run Step 2 |
Seedance | Tried Seedance instead of Kling | Use Kling only — see engine-choice section above |
Kling | Reference photo is a recognizable celebrity / public figure | Ask the user for a non-celebrity reference. Kling correctly blocks impersonation patterns (celebrity face + fake-event chyron) |
| Reference is an iPhone HEIC-derived JPEG with heavy EXIF and/or extreme aspect ratio (e.g. 2316×3088) | Re-encode the reference before upload: |
| gpt-image-2 high is a deliberately slower fidelity tier, not a bug — upstream typical is around two minutes per the manifest | Wait it out — most runs return cleanly. If a specific run does fail, retry once; fall back to |
| 症状 | 原因 | 修复方案 |
|---|---|---|
| 字幕条在视频中途弹出(约4–5秒闪烁) | 字幕条未嵌入静态图 | 重新执行步骤1;执行步骤2前确认 |
| 计分牌在视频中途出现动画/变形 | | 恢复严格匹配设置和完整的negative_prompt |
| 视频后期身份偏移(约10秒后面部变化) | 参考图片过小/Kling丢失面部特征 | 重新执行步骤2;如果偏移仍存在,重新执行步骤1并裁剪静态图的面部区域(面部像素越多,锁定效果越强) |
| 解说员读错用户名 | 原生音频为单次生成 | 重新执行步骤2 |
Seedance返回 | 使用了Seedance而非Kling | 仅使用Kling——参见上方引擎选择部分 |
Kling返回 | 参考图片为可识别的名人/公众人物 | 要求用户提供非名人参考图片。Kling会正确拦截冒充模式(名人面部+假事件字幕条) |
| 参考图片是带有大量EXIF信息或极端宽高比的iPhone HEIC转换JPEG(例如2316×3088) | 上传前重新编码参考图片: |
| gpt-image-2的high质量级别是故意设计的慢保真层级,并非bug——上游文档显示通常约需两分钟 | 等待即可——大多数运行会正常返回。如果特定运行失败,重试一次;仅在持续失败时 fallback 到 |
What NOT to do
禁止操作
- Don't sport-swap. NBA / NFL / soccer variants → fork this skill; don't parameterize this one.
- Don't add suffixes to the chyron (e.g. " - AI Creator"). Chyron is the username alone — the trend illusion depends on it reading like a real broadcast identifier.
- Don't add post-edits — no ,
add_captions,generate_music. Kling burns the scorebug + chyron + native commentary directly; anything added afterward breaks the broadcast illusion.edit_*
- 不要更换体育项目。NBA / NFL / 足球变体需复刻此技能;不要在此技能中添加参数化设置。
- 不要在字幕条中添加后缀(例如“ - AI创作者”)。字幕条仅显示用户名——特效的真实感依赖于其与真实转播标识一致。
- 不要添加后期编辑——不要使用 、
add_captions、generate_music。Kling已直接嵌入计分牌、字幕条和原生解说;后续添加任何内容都会破坏转播真实感。edit_*