模型未返回最终正文（finish_reason=length），以下为推理过程摘录，可尝试提高 max_tokens 或改用多段生成。

我们需要遵循用户要求。输出直接可写入数据库的 HTML 正文片段。禁止 markdown 代码围栏、禁止解释。允许的 HTML 标签仅有 <p>、<h3>。不要 img、script/style 等。需要按结构严格：第一个元素是 <p> 首段改写应用介绍，然后仅按小节顺序输出。小节顺序给出：亮点 → 优势 → 操作指南 → 推荐理由 → 特色 → 使用说明 → 测评。但注意到整体结构中说：已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者；若含「介绍」「简介」类关键词会固定排在首段之后的第一个 h3，其余小节再按其下列顺序；末位固定测评，全文共 4～7 个 h3。但随后明确列出了各小节任务：亮点、优势、操作指南、推荐理由、特色、使用说明、测评。这似乎有矛盾：到底输出全部7个小节还是随机抽取3~6个非测评小节? Let's parse carefully.

User starts: "【整体结构】顺序必须严格如下：
1) 第一个元素是一个 <p>...
2) 然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>... 非「测评」小节为 **若干** 个 <p> ... 每节要点条数以该小节下方【条数】为准 ...
3) **最后一个小节固定为「海绵音乐安卓版测评」**（已在下列顺序末位）...
【小节顺序】亮点 → 优势 → 操作指南 → 推荐理由 → 特色 → 使用说明 → 测评"

Then "【各小节任务与已代入模板】" includes all seven subsections, each with fixed title and number of items. That suggests perhaps they want all seven? But the "随机抽3~6个非测评小节" phrase indicates only some of the list should be selected? However, the instructions below explicitly define each subsection with 【条数】, making it seem all are required? Let's examine.

They say "已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）。" This is a prompt template generated by some system. It might have randomly selected a subset, but here the actual list includes all? The "各小节任务与已代入模板" contains all 7, but perhaps that's just enabling, and we need to infer there is no info about which subset selected. The "本节固定 3 条" etc. It says "若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>" but no subsection titled 介绍/简介 in the list; only "亮点". So that clause irrelevant.

Given "小节顺序" says "亮点 → 优势 → 操作指南 → 推荐理由 → 特色 → 使用说明 → 测评", and the user mandates "【输出硬性要求】" at end. It might be safest to output all seven h3? But "全文共 4～7 个 <h3>" means if all seven, there are 7 h3 (including last), which satisfies 4-7. It says "已从词池随机抽 3～6 个非测评小节" but the list below has 6 non-test subsections (亮点, 优势, 操作指南, 推荐理由, 特色, 使用说明). If all six selected, that's within 3-6. So maybe the random selection is exactly these six? There's no indication that some should be omitted; on the contrary, they include each subsection with fixed条数. Thus I should output all six non-test subsections plus one test = 7 h3. This aligns with the explicit per-subsection instructions. But the phrase "随机抽3～6个" could imply not all. However, because they provided detailed content for all 6, I think we need to include all. If we omitted any, we'd fail the per-section requirements that are explicit. Let's check: There is no "介绍" section, so no first H3 exception. So output:

First a <p> with rewritten intro. Must include app name naturally. Need based on original first paragraph: "海绵音乐app是字节跳动推出的一款完全免费的AI音乐创作工具，为用户提供个性化且多样化的音乐制作体验。该应用采用先进的深度学习模型与大数据分析技术，通过识别用户的音乐偏好、风格倾向及创作需求，智能生成相匹配的音乐内容。海绵音乐APP同时兼容多种音乐风格，涵盖R&B、" It ends with "涵盖R&B、" incomplete. We should rewrite with emphasis app store intro. Must include name "海绵音乐安卓版" or "海绵音乐安卓版"? The app name says "海绵音乐安卓版" in title. In first paragraph text: "应用名称（须在首段 <p> 正文中自然出现）：海绵音乐安卓版". So first p should contain "海绵音乐安卓版". We can say "海绵音乐安卓版是字节跳动推出的一款完全免费的AI音乐创作工具..." Good.

Need strict no markdown. Allowed HTML tags only p and h3. No script/style. Should not output html/body. Should not use code fences. Need use Chinese punctuation. "数字+全角句号．" means use "1．" etc. Note Japanese/Chinese fullwidth dot? "．" is fullwidth dot, okay. Need every line separate <p> with each item. Ensure no numbering outside.

Need small sections. Let's plan.

First p: Introduce app. Need "海绵音乐安卓版是字节跳动推出的一款完全免费的AI音乐创作工具。它依托先进的深度学习模型与大数据分析，能识别用户的音乐偏好、风格倾向和创作需求，智能生成个性化音乐内容，同时支持R&B等多种风格。" Should include the incomplete "R&B、" maybe add more styles? The original says "涵盖R&B、" then truncated. We can complete "R&B、流行、电子等" maybe okay? But "根据原文内容中改" must be highly relevant. Since original truncated at R&B, we can say "涵盖R&B等多种风格" to be safe. Better: "涵盖R&B、流行、电子等多种音乐风格" Might introduce new details? It's likely acceptable because original says "涵盖R&B、" and ellipsis. But to avoid hallucination, say "涵盖R&B等多种风格". But "等多种风格" okay. Need not mention "安卓版" in every h3? H3 titles fixed include it. Fine.

Then H3 "亮点" with 3 p each numbered. Must about core features/experience. Let's draft:

1．AI智能作词作曲：只需输入简单描述，即可快速生成完整音乐作品。
2．深度音乐偏好识别：通过学习用户风格，生成最贴合个人口味的旋律。
3．完全免费创作：无门槛使用全部核心创作功能，自由生成并保存作品。

Hmm need "每条一句、每条约50字内". okay.

Then H3 "优势" with 3 p:
1．依托字节跳动技术积累，音质与生成质量稳定可靠。
2．操作界面简洁易用，零基础用户也能快速上手创作。
3．多风格兼容覆盖R&B等曲风，满足不同场景创作需求。

Maybe "优势" should be performance/ease/ecosystem. ok.

Then H3 "操作指南" with 5 p. Need key flows:
1．安装后打开App，点击首页“开始创作”进入AI生成界面。
2．在文本框中输入主题、心情或风格关键词，描述想要的音乐感觉。
3．选择歌曲时长、曲风与配器偏好，点击生成等待AI输出结果。
4．生成后可在线试听，对不满意的段落点击“重做”或调整描述。
5．试听满意后保存到本地，并可为作品命名或设置专辑封面。

Need perhaps mention "海绵音乐安卓版操作指南" exactly.

Then H3 "推荐理由" with 5 p:
1．完全免费即可体验AI音乐创作，适合新手与业余爱好者。
2．无需乐理基础，输入文字就能获得旋律、和弦与编曲。
3．由字节跳动打造，算法持续更新，创作体验不断升级。
4．支持多种音乐风格，快速产出可作为背景音乐或灵感草稿。
5．手机端随时随地创作，灵感出现时立刻记录并生成音乐。

Need "落到具体收益" okay.

Then H3 "特色" with 6 p. Need distinct ability/scenes:
1．智能歌词与旋律同步生成，一键产出带人声演唱的完整歌曲。
2．支持用户对音乐偏好深度定制，AI会依据听歌记录调整风格。
3．提供R&B、流行、电子等多种曲库模板，覆盖多样化创作场景。
4．生成速度快，适合短视频配乐、个人Demo与创意热歌制作。
5．自带社区或分享功能？ Might be invented. Need be careful. We don't know. Better avoid uncertain features. Include only from original: preference, styles, free, deep learning. But need 6 distinct items. Can phrase:
1．深度学习模型可理解自然语言描述，将文字灵感迅速转化为音乐片段。
2．大数据分析用户偏好，让生成结果贴合每个人的独特审美。
3．支持R&B等曲风，并可根据需要调整节奏与情感氛围。
4．无需专业设备，手机端即可完成从创作到输出的音乐全流程。
5．生成内容可反复修改，帮助创作者不断打磨旋律与编曲。
6．完全免费使用，大幅降低音乐制作门槛，适合不同水平用户。

Maybe "情感氛围" is speculative but okay.

Then H3 "使用说明" with 3 p. Installation, permissions, account:
1．从应用商店下载安装后，注册并登录字节系账号即可开始使用。
2．首次使用可按需开启麦克风权限，用于语音输入创作灵感。
3．生成作品默认保存在个人作品库，可随时播放、编辑或导出。

Need "权限" maybe should not say microphone if not sure? Could be okay but perhaps "按需" suggests optional. Better to avoid specifics? The instruction says "安装、权限、账号与基础上手要点". We can include:
1．在官方应用商店下载海绵音乐安卓版，安装后按照引导完成基础设置。
2．使用手机号或抖音账号登录，即可同步创作数据并保存作品。
3．创作功能无需额外付费，但下载或分享时请留意网络与存储空间。

This is safer. But "下载或分享时请留意" is not permission. Maybe:
1．安装时需授权存储空间权限，以便保存生成的音乐文件。
2．建议登录账号后再创作，防止作品丢失并支持多设备同步。
3．无需Musical knowledge? Actually "使用说明" more operational. Let's do:
1．安装后首次启动可根据引导开启通知权限，方便接收创作草稿提醒。
2．使用手机号或第三方账号登录，作品将自动缓存至个人创作空间。
3．点击右上角“帮助与设置”可管理音质、存储与智能生成参数。

Hmm maybe "音质" unknown. Avoid.

Use 3 p:
1．下载安装后建议先完成登录，以便将作品保存到云端创作空间。
2．首次使用可按需授权存储和麦克风权限，便于录制灵感与导出音频。
3．生成内容前可先浏览“创作指南”，了解风格标签与操作快捷键。

But "快捷键" inappropriate for mobile. Let's design more generic.

Since this is a generated prompt, we can fill with reasonable content. Need "各小节任务" says "本小节无对应采集原文，按「原创H3-普通」撰写". So we can invent plausible.

Need final H3 "海绵音乐安卓版测评" and one <p> no numbering, <=200 chars. We need combine features/adv: "海绵音乐安卓版测评" is last h3. Then a short paragraph.

Let's ensure total "<h3>" count = 7. Or possibly "全文共 4～7 个 <h3>" yes 7 is okay. But if "random selection 3-6 non-test subsections" within, 6 non-test selected. Good.

Need no markdown fences. We should output raw HTML. Need maintain exact titles: <h3>海绵音乐安卓版亮点</h3>, etc. For the last, <h3>海绵音乐安卓版测评</h3>. Ensure no trailing/final comments.

Need first p is first element. Then each h3 after p. There should be no leading p? Actually first element is p. OK.

Need spacing/newlines? It's acceptable. Use raw HTML.

Need ensure each numbered item is within a separate p. "每条单独成行" so each <p> contains number and content. That's good.

Need "全篇统一为同一种序号格式" — use "1．", not "1.". So each p begins with "1．", "2．" etc. Use fullwidth Japanese dot. Good.

Need "每节要点条数以该小节下方【条数】为准". So 3/3/5/5/6/3. We must count.

Need "最后一个小节固定为「海绵音乐安卓版测评」：仅 1 个 p，不要序号与分条，一段短评，全文不超过200字". So the final p after H3 has no number.

Need "全文不超过200字" for final only? It says "全文不超过 200 字（含标点）" within测评. Yes.

Need "禁止输出 <img>（应用截图由系统在生成后按小节顺序自动插入，模型无需处理图片）" No images.

Need "不要输出 <html><body> 等外壳" ok.

Now potential concern: "首段模板（{head}/{text} 均已代入为首段原文）改写应用介绍" We need "与原文的表达主旨保持高度关联与一致性... 内容必须包含游戏/软件名称... 标题：海绵音乐安卓版；原文首段..." We should not use "原文首段" verbatim? We need original rewrite. We'll make it prose with app name in first sentence. Should not transcribe the incomplete "涵盖R&B、" with comma, but incorporate.

Let's construct first p carefully:

"海绵音乐安卓版是字节跳动推出的一款完全免费的AI音乐创作工具，致力于让每个人都能轻松制作个性化音乐。它借助深度学习模型与大数据分析，精准识别用户的音乐偏好与创作需求，智能生成风格匹配的音乐内容，并支持R&B、流行、电子等多种曲风。"

This includes "R&B、流行、电子" maybe invented but plausible. To be safe "多种曲风" without listing? "支持R&B、流行、电子等多种曲风" is okay. But "原文首段" says "涵盖R&B、" truncated. Maybe "涵盖R&B等多种风格" no new styles. Let's use "涵盖R&B、流行、电子等多种风格" because it's natural and user asked to rewrite, not necessarily avoid extending. But if strict "根据原文" perhaps no issue. We'll decide.

Need first p can be more than one sentence. Good.

Now write each subsection. Let's think about "亮点" "优势" "特色" distinctions. Need avoid overlapping too much.

Detailed draft:

<p>海绵音乐安卓版是字节跳动推出的一款完全免费的AI音乐创作工具，致力于让每个人都能轻松制作个性化音乐。它借助深度学习模型与大数据分析，精准识别用户的音乐偏好与创作需求，智能生成风格匹配的音乐内容，涵盖R&B、流行、电子等多种曲风。</p>

<h3>海绵音乐安卓版亮点</h3>
<p>1．AI智能创作：输入简单描述即可快速生成旋律、歌词与编曲。</p>
<p>2．个性化推荐：根据你的音乐偏好和风格倾向定制专属作品。</p>
<p>3．完全免费：核心创作功能零门槛开放，随时产出完整音乐。</p>

Need each under 50 chars? Let's count: 1) Chinese ~? "AI智能创作：输入简单描述即可快速生成旋律、歌词与编曲。" 24? fine.

<h3>海绵音乐安卓版优势</h3>
<p>1．依托字节跳动的算法实力，音乐生成质量稳定且持续进化。</p>
<p>2．操作界面清晰简洁，零基础用户也能快速进入创作流程。</p>
<p>3．兼容R&B、流行、电子等曲风，创作场景覆盖日常与专业需求。</p>

Need "优势" maybe "性能、易用性、兼容性" — yes.

<h3>海绵音乐安卓版操作指南</h3>
<p>1．下载并安装海绵音乐安卓版，打开后点击首页的“创作”按钮。</p>
<p>2．在输入框中填写主题、心情或风格关键词，描述你想要的音乐方向。</p>
<p>3．选择歌曲时长与参考曲风，也可以保留默认参数直接生成。</p>
<p>4．生成后在线试听，若效果不理想可点击“重新生成”或修改描述。</p>
<p>5．满意后保存至作品库，支持继续编辑歌词、封面与整体混音。</p>

Potential "参考曲风" maybe okay. But "不满意的段落点击重做" maybe not. Fine.

<h3>海绵音乐安卓版推荐理由</h3>
<p>1．完全免费让音乐创作不再受成本限制，适合所有入门新手。</p>
<p>2．无需乐理基础，一句话就能得到旋律、和弦和完整编曲。</p>
<p>3．生成速度快，适合短视频创作者快速产出背景音乐。</p>
<p>4．支持个性化偏好识别，每次生成都更贴近你的听歌口味。</p>
<p>5．手机端操作便捷，灵感出现时能随时随地把想法变成作品。</p>

Need "和弦" "完整编曲" based on original maybe "音乐制作体验" but not likely? It's an AI tool. Fine.

<h3>海绵音乐安卓版特色</h3>
<p>1．深度学习模型能将自然语言描述迅速转化为可听的音乐片段。</p>
<p>2．大数据分析个人听歌记录，让生成结果贴合用户独特审美。</p>
<p>3．覆盖R&B、流行、电子等风格，并根据心情词调整节奏气氛。</p>
<p>4．文字描述即可驱动人声演唱与歌词生成，降低创作门槛。</p>
<p>5．支持边听边改，持续调节旋律、配器和情绪，打造个人作品。</p>
<p>6．从灵感到成曲均在手机端完成，满足日常记录、练歌与配乐需求。</p>

Potential "人声演唱" not in original but could be an AI music app. We should avoid making up features not original? The instruction says "原文首段" only mentions generating music content and style. But "特色" section "相对同类软件最突出的差异能力与使用场景" can have plausible. We need keep only "根据原文中的内容" but no original for this subsection; "本小节无对应采集原文，按「原创H3-普通」撰写" means we create based on original. So "人声演唱" is plausible but could be unsupported. Maybe safer: "完整歌曲" not "人声". Let's adjust:

1．深度学习模型覆盖从歌词到编曲的完整创作链路，生成内容自然连贯。
2．大数据分析个人偏好，智能匹配用户喜欢的节奏与风格。
3．兼容R&B、流行、电子等多种风格，可依据心情关键词切换氛围。
4．文本输入即可驱动旋律生成，免去乐理和乐器操作门槛。
5．生成结果支持反复调整，方便在草稿基础上打磨出完整单曲。
6．手机端一站式完成灵感捕捉、音乐生成与作品存储。

Good.

<h3>海绵音乐安卓版使用说明</h3>
<p>1．在海绵音乐安卓版应用商店页面下载安装，首次启动按引导完成基础设置。</p>
<p>2．建议使用手机号或第三方账号登录，生成的作品可自动保存在个人空间。</p>
<p>3．使用时可开启存储权限用于导出音频，并在设置中调整生成参数。</p>

Need "第三方账号登录" maybe okay. "生成参数" maybe plausible.

<h3>海绵音乐安卓版测评</h3>
<p>海绵音乐安卓版把专业音乐制作的门槛降到了“一句话”级别。整个创作过程流畅直观，AI生成的旋律与编曲自然度超出预期，而且完全免费。对于不会乐理但想表达灵感的普通人来说，它既是随手记旋律的创作本，也是快速产出配乐的效率工具。偏好识别和多风格支持让每次生成都有新鲜感，稍加调整就能变成属于自己的完整作品。</p>

Count final p? Let's count Chinese chars ~ maybe 156. Let's count exactly to be safe. It should be <=200. Use approximate. Let's count string:

"海绵音乐安卓版把专业音乐制作的门槛降到了“一句话”级别。" 
海1 绵2 音3 乐4 安5 卓6 版7 把8 专9 业10 音11 乐12 制13 作14 的15 门16 槛17 降18 到19 了20 “21 一22 句23 话24 ”25 级26 别27 。28
"整个创作过程流畅直观，" 整个29 创30 作31 过32 程33 流34 畅35 直36 观37 ，38
"AI生成的旋律与编曲自然度超出预期，而且完全免费。" AI 39-40? Let's treat "AI" 2 chars: 生41 成42 的43 旋44 律45 与46 编47 曲48 自49 然50 度51 超52 出53 预54 期55 ，56 而57 且58 完59 全60 免61 费62 。63
"对于不会乐理但想表达灵感的普通人来说，" 对64 于65 不66 会67 乐68 理69 但70 想71 表72 达73 灵74 感75 的76 普77 通78 人79 来80 说81 ，82
"它既是随手记旋律的创作本，也是快速产出配乐的效率工具。" 它83 既84 是85 随86 手87 记88 旋89 律90 的91 创92 作93 本94 ，95 也96 是97 快98 速99 产100 出101 配102 乐103 的104 效105 率106 工107 具108 。109
"偏好识别和多风格支持让每次生成都有新鲜感，" 偏110 好111 识112 别113 和114 多115 风116 格117 支118 持119 让120 每121 次122 生123 成124 都125 有126 新127 鲜128 感129 ，130
"稍加调整就能变成属于自己的完整作品。" 稍131 加132 调133 整134 就135 能136 变137 成138 属139 于140 自141 己142 的143 完144 整145 作146 品147 。148

~148. Good.

Need "语气客观简练" okay.

Now, however, the prompt says "全文共 4～7 个 <h3>". We include 7. Need perhaps "若含「介绍」「简介」类关键词会固定排在首段之后的第一个 <h3>" not relevant. But if we include all six, maybe there is no random selection. But instructions from "各小节任务" explicitly require all six, so output all. Good.

Potential issue: "非「测评」小节为 若干 个 <p>、每条要点须带序号且全篇统一为同一种序号格式（本次要求：用序号分条：每条单独成行，行首为「数字+全角句号．」（示例 1．2．3．），全篇保持同一格式）" We will use exactly.

Need "每条单独成行" each p is line. OK.

Need "每节要点条数以该小节下方【条数】为准" yes.

Need "首段 <p> 正文中自然出现" our first p has "海绵音乐安卓版". Good.

Need "标题：海绵音乐安卓版；原文首段：海绵音乐app是...". We might include "海绵音乐安卓版" in first p. Good.

Potential issue: "原文首段" ends with "涵盖R&B、" but not closed. It may be a placeholder from tuncated source. We should not include the dangling comma. Fine.

Need "依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。" We did.

Need "强制 HTML tags only p,h3" We need ensure no stray text outside tags. Newlines are okay. But no text outside tags. We'll output exactly.

Could "禁止 markdown 代码围栏" means no triple backticks. We'll not.

Need "不要输出 <html><body> 等外壳，不要 script/style." OK.

Need perhaps "应用截图由系统在生成后按小节顺序自动插入，模型无需处理图片" Not relevant.

Now, let's consider if "小节顺序" says "亮点 → 优势 → 操作指南 → 推荐理由 → 特色 → 使用说明 → 测评" But "操作指南" after "优势" before "推荐理由" yes. But if all six non-test, order: 亮点, 优势, 操作指南, 推荐理由, 特色, 使用说明, 测评. Note that the list in "整体结构" is "亮点 → 优势 → 操作指南 → 推荐理由 → 特色 → 使用说明 → 测评". yes. We'll follow.

Now maybe "已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" means we should maybe select a subset? Let's examine if there are six non-test subsections in "各小节任务". It may be because the template includes all subsections but in actual generation "已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" is a cue to only output some? But then "最终输出" may depend on "小节任务" which appears for every possible section, not necessarily selected. The user then says "每个小节先 <h3>（标题须与下列 {H3title} 完全一致）" Hmm "下列" refers to "【小节顺序】"? Let's reread:

"2) 然后**仅按下列「小节顺序」**输出（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+全角句号．」（示例 1．2．3．），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。"

Then "【小节顺序】亮点 → 优势 → 操作指南 → 推荐理由 → 特色 → 使用说明 → 测评"

The "下列 {H3title}" maybe "下列" is the list of key names? It says "与下列 {H3title} 完全一致" but {H3title} has not been shown except in each subsection. Could be from template. But in final instructions, they included all subsection headings. I think we should produce all.

Maybe there is a hidden instruction "首段模板 (head/text)" substitutes all variables; "小节任务" includes explicit "本小节无对应采集原文". This is generated by some pipeline that has decided not to include original text for any section. It also says "从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" but no original text is provided for any non-test subsection, so maybe it should randomly choose? Yet "各小节任务" describes all seven. The model could be expected to output all of them, because "各小节任务" lists specific条数 for all sections. However, "全文共 4～7 个 <h3>" if output all 7 is okay. If output only 5, still okay but would need to omit some sections. But the instruction "必须严格如下" and "仅按下列小节顺序输出" implies all sections in the list? The "随机抽" is contradictory. Let's see if there is "优先保留采集中有原文者" — maybe "采集中有原文者" could refer to the provided "原文首段" only, not other sections. The "若含「介绍」「简介」类关键词会固定排在..." maybe from a broader template. Could be irrelevant.

Potentially, to be safe, we can include all six non-test sections. If the grader expected fewer, including extra may violate "从词池随机抽 3~6" but since 6 is within range, it's okay. The "全文共 4~7" includes 7. So okay.

Need maybe not include "<p>" number lines with single newlines? We can have newline between tags. Good.

Now let's think about content quality and originality. Need "原创改写" not copied verbatim. We used phrase "完全免费" from original. OK.

Need "应用商店的介绍文案" tone, so first p should be engaging. Good.

Let's check each subsection's "每条约50字内" — ours are about 20-30 chars. Good.

Potential issue: H3 title "海绵音乐安卓版亮点" vs "海绵音乐安卓版亮点" exactly? The template says "标题必须是 <h3>海绵音乐安卓版亮点</h3>". We need exactly. Ensure "安卓版" not "APP".

Now final p "不超过200字" we checked.

Need "不要使用 1、2、3" in final, we don't.

Now let's draft final response with raw HTML.

Before final, let's review if any "script/style" accidentally? No.

Let's consider "允许的 HTML 标签仅有：<p>、<h3>。" We use no attributes.

Potential "代码围栏" no.

Potential "不需要处理图片" no img.

Now think about "首段 <p> 正文中自然出现：海绵音乐安卓版" first p contains "海绵音乐安卓版" at start. Good.

Need "title: 海绵音乐安卓版" maybe not output title tag. Fine.

Now let's refine content to avoid "游戏" or "软件" weirdness. It's an app, not a game. Our copy doesn't call it game. Good.

Potential "字节跳动" capitalization: "字节跳动" ok.

Need "多种音乐风格，涵盖R&B、流行、电子等" good.

Let's ensure all numbers format "1．" not "1." exactly. We'll use "1．" etc.

Let's check each section count:

- Intro p: 1
- H3 亮点 + 3 p
- H3 优势 + 3 p
- H3 操作指南 + 5 p
- H3 推荐理由 + 5 p
- H3 特色 + 6 p
- H3 使用说明 + 3 p
- H3 测评 + 1 p

Total h3 7, total p 1 + 3+3+5+5+6+3+1 = 27. Good.

Need maybe "每节要点条数以该小节下方【条数】为准（在 3～6 条之间且各节独立）" Our counts: 3,3,5,5,6,3. all between 3-6. Good.

Now, let's think if "操作指南" should be "关键流程的分步说明" and "使用说明" "安装、权限、账号与基础上手要点" to avoid overlap. Our "操作指南" starts with "下载并安装", which overlaps "使用说明" installation. We can revise to avoid redundant. "使用说明" installation; "操作指南" first step could start after opening. But it's okay; "操作指南" can begin with open app not install. Let's adjust:

操作指南:
1．打开海绵音乐安卓版，点击底部“创作”进入AI音乐生成页面。
2．在文本区填写主题、心情或风格关键词，描述想要的音乐方向。
3．选择曲风、时长等偏好，或直接用默认参数开始生成。
4．生成后在线试听，针对不理想片段可点击“重做”或修改描述。
5．对结果满意后保存至作品库，并继续编辑歌词、封面等信息。

"使用说明" installation:
1．在应用商店搜索并下载海绵音乐安卓版，安装后按引导完成启动设置。
2．使用手机号或支持的第三方账号登录，作品会自动保存在个人云端空间。
3．首次使用按需开启存储权限，以便导出音频及管理创作文件。

This reduces overlap. "点底部创作" maybe interface unknown but okay. "底部" maybe not certain; avoid.

Operation guide:
1．打开海绵音乐安卓版，点击首页核心入口“创作”进入生成页。
2．在文本框中输入主题、心情或风格关键词，描述想要的音乐方向。
3．按需选择曲风、时长等偏好，也可保留默认参数直接生成。
4．生成后在线试听，若结果不满意可点击“重新生成”或修改描述。
5．确认效果后保存到个人作品库，支持继续编辑作品信息。

Good.

"使用说明" with permissions/account:
1．在应用商店下载安装海绵音乐安卓版，首次启动按引导授予必要权限。
2．建议使用手机号或第三方账号登录，创作内容将自动保存至云端空间。
3．可在设置中管理导出音质与存储位置，并随时查看设备存储空间。

"导出音质与存储位置" maybe invented but plausible. Better:
3．如需导出到相册或本地，请确保已开启存储权限并预留足够空间。

Good.

Need "首次启动按引导授予必要权限" maybe vague but okay.

Now "特色" section 6. Let's craft distinct:

1．深度学习模型能够理解自然语言描述，快速将文字灵感转为音乐片段。
2．通过大数据分析用户偏好，精准匹配节奏、旋律与情绪氛围。
3．支持R&B、流行、电子等多种风格，同样一段话可生成不同版本。
4．零乐理基础也可直接创作，消除乐器演奏与编曲知识门槛。
5．生成结果支持循环修改，持续调整直到接近创作者心中效果。
6．手机端完整覆盖从灵感到成品的创作流程，适合随时记录想法。

Good. Each ~30 chars.

Need "推荐理由" 5:
1．完全免费即可使用核心AI创作功能，对新手十分友好。
2．无需乐理或编曲经验，输入文字就能获得完整音乐片段。
3．字节跳动技术加持，生成质量与稳定性更有保障。
4．多风格与偏好识别让作品更贴近个人口味，提升创作乐趣。
5．随手用手机就能创作，适合短视频配乐、个人Demo等实际场景。

Good.

Need "优势" 3:
1．依托字节跳动深度学习技术，生成内容自然且风格统一。
2．操作流畅、上手简单，从输入描述到得到作品只需几步。
3．兼容R&B、流行、电子等多样曲风，覆盖不同创作需求。

Need "亮点" 3:
1．AI一键生成完整音乐片段，让创作过程更高效。
2．根据用户偏好与风格倾向智能定制，每首作品都有个性化特色。
3．完全免费开放核心功能，降低音乐创作门槛。

Good.

Let's think about first p: It says "海绵音乐安卓版是字节跳动推出的一款完全免费的AI音乐创作工具，致力于让每个人都能轻松制作个性化音乐。它借助深度学习模型与大数据分析，精准识别用户的音乐偏好与创作需求，智能生成风格匹配的音乐内容，涵盖R&B、流行、电子等多种曲风。" This is okay. Maybe "涵盖R&B、流行、电子等多种曲风" no "及" issue.

Potential "原文首段" says "海绵音乐APP同时兼容多种音乐风格，涵盖R&B、" In original "同时兼容多种音乐风格" and "涵盖R&B、" not "流行、电子". We should maybe say "兼容R&B等多音乐风格" to avoid inventing. But "流行、电子" is an obvious extrapolation from "涵盖R&B、" (which means there are more). Acceptable.

Need "原文的表达主旨保持高度关联与一致性" yes.

Now final paragraph: "海绵音乐安卓版把专业音乐制作的门槛降到了“一句话”级别。整个创作过程流畅直观，AI生成的旋律与编曲自然度超出预期，而且完全免费。对于不会乐理但想表达灵感的普通人来说，它既是随手记旋律的创作本，也是快速产出配乐的效率工具。偏好