<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>Pengliang Li</title>
    <link>https://rant.li/pengliangli/</link>
    <description></description>
    <pubDate>Wed, 18 Nov 2026 23:05:45 +0000</pubDate>
    <item>
      <title>Turning a selfie into a short fan-made Sheru Bhai reel</title>
      <link>https://rant.li/pengliangli/turning-a-selfie-into-a-short-fan-made-sheru-bhai-reel</link>
      <description>&lt;![CDATA[I wanted a short vertical clip I could add captions to later, starting from one photo I actually have permission to use. I did not want a subscription, and I also did not want a tool that pretends the character is a real person or that a meme fortune is a documented fact.&#xA;&#xA;The loop I use is straightforward:&#xA;&#xA;Start with the free, watermarked still, which does not require an account, to see whether the look is even close to what I want.&#xA;&#xA;For a clip, pick one of the four original scene directions — a luxury-car entrance, a comic auction, a throne portrait, or Walk Beside Sheru Bhai — or use a clear adult selfie I have permission to use.&#xA;&#xA;Generate an approximately five-second, silent 9:16 MP4, then add captions or licensed audio in an editor before sharing.&#xA;&#xA;Review the result carefully. It is fan-made AI entertainment, not a real likeness and not an official clip.&#xA;&#xA;I have been using Sheru Bhai AI Video Generator for this. It is an independent, fan-made site for short vertical clips with the fictional AI lion character Sheru Bhai. It is not the official Sheru Bhai site and is not affiliated with @digitalsheru; it generates its own pictures and clips rather than using the original creator’s videos or competitor artwork. The $999 trillion figure that shows up in the joke auction scene is a meme, not a verified fortune. Video generation uses one-time credit packs rather than a subscription, unused credits do not expire, and a failed or timed-out render returns its credit. It will not replace a real video shoot, but it is a practical way to draft a short, silent reel I can edit before posting.]]&gt;</description>
      <content:encoded><![CDATA[<p>I wanted a short vertical clip I could add captions to later, starting from one photo I actually have permission to use. I did not want a subscription, and I also did not want a tool that pretends the character is a real person or that a meme fortune is a documented fact.</p>

<p>The loop I use is straightforward:</p>
<ul><li><p>Start with the free, watermarked still, which does not require an account, to see whether the look is even close to what I want.</p></li>

<li><p>For a clip, pick one of the four original scene directions — a luxury-car entrance, a comic auction, a throne portrait, or Walk Beside Sheru Bhai — or use a clear adult selfie I have permission to use.</p></li>

<li><p>Generate an approximately five-second, silent 9:16 MP4, then add captions or licensed audio in an editor before sharing.</p></li>

<li><p>Review the result carefully. It is fan-made AI entertainment, not a real likeness and not an official clip.</p></li></ul>

<p>I have been using <a href="https://sherubhai.pro/" rel="nofollow">Sheru Bhai AI Video Generator</a> for this. It is an independent, fan-made site for short vertical clips with the fictional AI lion character Sheru Bhai. It is not the official Sheru Bhai site and is not affiliated with @digitalsheru; it generates its own pictures and clips rather than using the original creator’s videos or competitor artwork. The $999 trillion figure that shows up in the joke auction scene is a meme, not a verified fortune. Video generation uses one-time credit packs rather than a subscription, unused credits do not expire, and a failed or timed-out render returns its credit. It will not replace a real video shoot, but it is a practical way to draft a short, silent reel I can edit before posting.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/turning-a-selfie-into-a-short-fan-made-sheru-bhai-reel</guid>
      <pubDate>Fri, 13 Nov 2026 10:44:37 +0000</pubDate>
    </item>
    <item>
      <title>Studying TCM herbs and formulas with source labels in two languages</title>
      <link>https://rant.li/pengliangli/studying-tcm-herbs-and-formulas-with-source-labels-in-two-languages</link>
      <description>&lt;![CDATA[When I study Traditional Chinese Medicine, I often move between English and Chinese terms for herbs, formulas, and pattern theory. The hard part is not finding an answer; it is keeping classical sources, later TCM interpretation, and modern research from collapsing into one unqualified claim.&#xA;&#xA;The habit I try to keep:&#xA;&#xA;Start with a focused question — about a herb, a formula, an acupoint, or a pattern — rather than a vague symptom dump.&#xA;Ask for the related terms and how they show up in classical texts versus later interpretation.&#xA;Check whether a source is labeled (classical source, TCM interpretation, modern research, or safety) instead of treating every citation as the same kind of evidence.&#xA;Stop there. This is study, not a diagnosis or a treatment plan.&#xA;&#xA;I have been using TCM Sage as a bilingual TCM AI companion for this kind of reading. It is designed for careful study of herbs, formulas, classics, channels, acupoints, and pattern theory, and it aims to show sources when they are available rather than invent citations. It is educational only and is not a substitute for medical care.]]&gt;</description>
      <content:encoded><![CDATA[<p>When I study Traditional Chinese Medicine, I often move between English and Chinese terms for herbs, formulas, and pattern theory. The hard part is not finding an answer; it is keeping classical sources, later TCM interpretation, and modern research from collapsing into one unqualified claim.</p>

<p>The habit I try to keep:</p>
<ul><li>Start with a focused question — about a herb, a formula, an acupoint, or a pattern — rather than a vague symptom dump.</li>
<li>Ask for the related terms and how they show up in classical texts versus later interpretation.</li>
<li>Check whether a source is labeled (classical source, TCM interpretation, modern research, or safety) instead of treating every citation as the same kind of evidence.</li>
<li>Stop there. This is study, not a diagnosis or a treatment plan.</li></ul>

<p>I have been using <a href="https://tcmsage.com" rel="nofollow">TCM Sage</a> as a bilingual TCM AI companion for this kind of reading. It is designed for careful study of herbs, formulas, classics, channels, acupoints, and pattern theory, and it aims to show sources when they are available rather than invent citations. It is educational only and is not a substitute for medical care.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/studying-tcm-herbs-and-formulas-with-source-labels-in-two-languages</guid>
      <pubDate>Fri, 13 Nov 2026 05:07:28 +0000</pubDate>
    </item>
    <item>
      <title>Turning a text description into a song demo I can preview in the browser</title>
      <link>https://rant.li/pengliangli/turning-a-text-description-into-a-song-demo-i-can-preview-in-the-browser</link>
      <description>&lt;![CDATA[I often have a mood, a lyric fragment, or a genre in mind before I have a project file in a DAW. Opening a full production setup just to test whether an idea is even worth arranging is more friction than I want at that stage.&#xA;&#xA;The loop I use:&#xA;&#xA;Write one description of the sound, story, mood, or lyric idea.&#xA;Set a style (such as Hip-Hop, Pop, Jazz, or EDM), a mood, instruments, and whether I want vocals or an instrumental.&#xA;Generate a complete track and preview it in the browser.&#xA;Keep the take if the direction is right, or adjust the description and try again.&#xA;Only then decide whether the idea is worth opening in a DAW.&#xA;&#xA;I have been using Seedmusic AI for this. It is a browser text-to-song generator with simple and custom modes, and it starts with free credits after signup. It will not replace a production session, but it is useful when I want a complete demo I can hear immediately.]]&gt;</description>
      <content:encoded><![CDATA[<p>I often have a mood, a lyric fragment, or a genre in mind before I have a project file in a DAW. Opening a full production setup just to test whether an idea is even worth arranging is more friction than I want at that stage.</p>

<p>The loop I use:</p>
<ul><li>Write one description of the sound, story, mood, or lyric idea.</li>
<li>Set a style (such as Hip-Hop, Pop, Jazz, or EDM), a mood, instruments, and whether I want vocals or an instrumental.</li>
<li>Generate a complete track and preview it in the browser.</li>
<li>Keep the take if the direction is right, or adjust the description and try again.</li>
<li>Only then decide whether the idea is worth opening in a DAW.</li></ul>

<p>I have been using <a href="https://seedmusic-ai.app" rel="nofollow">Seedmusic AI</a> for this. It is a browser text-to-song generator with simple and custom modes, and it starts with free credits after signup. It will not replace a production session, but it is useful when I want a complete demo I can hear immediately.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/turning-a-text-description-into-a-song-demo-i-can-preview-in-the-browser</guid>
      <pubDate>Fri, 13 Nov 2026 05:07:19 +0000</pubDate>
    </item>
    <item>
      <title>Generating nail-art reference images from a short style prompt</title>
      <link>https://rant.li/pengliangli/generating-nail-art-reference-images-from-a-short-style-prompt</link>
      <description>&lt;![CDATA[When I want manicure ideas, I usually need a few visual references before I compare colors or talk through a design. I do not want to open a full illustration tool or spend time masking photos of hands just to get a moodboard started.&#xA;&#xA;The loop I use:&#xA;&#xA;Write a short description of the style and color I have in mind.&#xA;Generate a few nail-art visuals and keep only the ones that match the mood.&#xA;Compare styles and colors side by side, then save the ones worth showing later.&#xA;Treat the images as inspiration, not as a finished appointment plan.&#xA;&#xA;I have been using NailsGen for this. It is a Flux Kontext-powered AI nail art generator that turns a prompt into manicure design visuals in seconds, in different styles and colors, and it does not require design skills. The current product is positioned as a free generator. It will not replace a nail technician, but it is a fast way to collect style references.]]&gt;</description>
      <content:encoded><![CDATA[<p>When I want manicure ideas, I usually need a few visual references before I compare colors or talk through a design. I do not want to open a full illustration tool or spend time masking photos of hands just to get a moodboard started.</p>

<p>The loop I use:</p>
<ul><li>Write a short description of the style and color I have in mind.</li>
<li>Generate a few nail-art visuals and keep only the ones that match the mood.</li>
<li>Compare styles and colors side by side, then save the ones worth showing later.</li>
<li>Treat the images as inspiration, not as a finished appointment plan.</li></ul>

<p>I have been using <a href="https://www.nailsgen.com" rel="nofollow">NailsGen</a> for this. It is a Flux Kontext-powered AI nail art generator that turns a prompt into manicure design visuals in seconds, in different styles and colors, and it does not require design skills. The current product is positioned as a free generator. It will not replace a nail technician, but it is a fast way to collect style references.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/generating-nail-art-reference-images-from-a-short-style-prompt</guid>
      <pubDate>Fri, 13 Nov 2026 05:07:11 +0000</pubDate>
    </item>
    <item>
      <title>Opening a spoiler-aware jubensha recap only after the script is finished</title>
      <link>https://rant.li/pengliangli/opening-a-spoiler-aware-jubensha-recap-only-after-the-script-is-finished</link>
      <description>&lt;![CDATA[After a 剧本杀 (script-murder) session, the table often keeps arguing about the timeline, a character&#39;s motive, or whether a clue was actually on the table. I want a place to look that up, but only after we have finished playing. A recap opened too early ruins the night.&#xA;&#xA;The habit I keep:&#xA;&#xA;Finish the script first. Do not open a truth breakdown during play.&#xA;Find the title by theme, player count, or duration, rather than by guessing a recap URL.&#xA;Use the dedicated review page for truth analysis, character relationships, timelines, and key clues.&#xA;Treat leftover placeholder pages as unfinished, not as a complete answer.&#xA;&#xA;I have been using Jubensha Review Navigator for this. It is a spoiler-aware directory and review hub for Chinese jubensha scripts, with structured entries and recap entry points. Content is meant for players who have already finished a script. It will not play the game for you, and some pages may still be waiting for a full recap, but it is useful when the group wants a shared reference after the session ends.]]&gt;</description>
      <content:encoded><![CDATA[<p>After a 剧本杀 (script-murder) session, the table often keeps arguing about the timeline, a character&#39;s motive, or whether a clue was actually on the table. I want a place to look that up, but only after we have finished playing. A recap opened too early ruins the night.</p>

<p>The habit I keep:</p>
<ul><li>Finish the script first. Do not open a truth breakdown during play.</li>
<li>Find the title by theme, player count, or duration, rather than by guessing a recap URL.</li>
<li>Use the dedicated review page for truth analysis, character relationships, timelines, and key clues.</li>
<li>Treat leftover placeholder pages as unfinished, not as a complete answer.</li></ul>

<p>I have been using <a href="https://jubensha.pro" rel="nofollow">Jubensha Review Navigator</a> for this. It is a spoiler-aware directory and review hub for Chinese jubensha scripts, with structured entries and recap entry points. Content is meant for players who have already finished a script. It will not play the game for you, and some pages may still be waiting for a full recap, but it is useful when the group wants a shared reference after the session ends.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/opening-a-spoiler-aware-jubensha-recap-only-after-the-script-is-finished</guid>
      <pubDate>Fri, 13 Nov 2026 05:07:01 +0000</pubDate>
    </item>
    <item>
      <title>Solving a cat logic puzzle with one piece per color, row, and column</title>
      <link>https://rant.li/pengliangli/solving-a-cat-logic-puzzle-with-one-piece-per-color-row-and-column</link>
      <description>&lt;![CDATA[I wanted a short logic puzzle I could open in a browser tab, without installing anything and without a competitive timer hanging over me.&#xA;&#xA;The rules I play with are simple:&#xA;&#xA;Place cats so every color, every row, and every column contains exactly one cat.&#xA;Turn on the optional no-touching constraint when I want a tighter board.&#xA;Use coordinates and hints only after I have already tried a placement out loud.&#xA;Switch to night mode or colorblind mode when the default palette is harder to read.&#xA;On a later run, swap the default cat piece for a custom image or GIF just to change how the board looks.&#xA;&#xA;I have been playing Cat Puzzle Game. It is a free browser logic puzzle with tutorial levels, scoring, lives, and a multilingual UI. No install is required. It is not a deep competitive puzzle platform, but it is a clean five-minute break when I want a constraint-satisfaction problem with a lighter theme.]]&gt;</description>
      <content:encoded><![CDATA[<p>I wanted a short logic puzzle I could open in a browser tab, without installing anything and without a competitive timer hanging over me.</p>

<p>The rules I play with are simple:</p>
<ul><li>Place cats so every color, every row, and every column contains exactly one cat.</li>
<li>Turn on the optional no-touching constraint when I want a tighter board.</li>
<li>Use coordinates and hints only after I have already tried a placement out loud.</li>
<li>Switch to night mode or colorblind mode when the default palette is harder to read.</li>
<li>On a later run, swap the default cat piece for a custom image or GIF just to change how the board looks.</li></ul>

<p>I have been playing <a href="https://catpuzzlegame.com" rel="nofollow">Cat Puzzle Game</a>. It is a free browser logic puzzle with tutorial levels, scoring, lives, and a multilingual UI. No install is required. It is not a deep competitive puzzle platform, but it is a clean five-minute break when I want a constraint-satisfaction problem with a lighter theme.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/solving-a-cat-logic-puzzle-with-one-piece-per-color-row-and-column</guid>
      <pubDate>Fri, 13 Nov 2026 05:06:51 +0000</pubDate>
    </item>
    <item>
      <title>Keeping DeepSeek prompts in a library instead of old chat threads</title>
      <link>https://rant.li/pengliangli/keeping-deepseek-prompts-in-a-library-instead-of-old-chat-threads</link>
      <description>&lt;![CDATA[When a prompt finally works well in DeepSeek, I used to leave it buried in an old chat. The next time I needed the same pattern — a writing rewrite, a coding task, or a creative brief — I would scroll, copy fragments, and slightly change the wording without remembering why the original worked.&#xA;&#xA;The loop I use now looks like this:&#xA;&#xA;Start from a ready-made template that is already close to the task, instead of a blank prompt.&#xA;Adjust the constraints in a management workspace so the prompt is reusable, not tied to one conversation.&#xA;Run the wording through a prompt optimizer when the output is close but inconsistent.&#xA;Keep the saved version somewhere I can open quickly, including from the browser.&#xA;&#xA;I have been using DeepSeek Prompt for this. It is a prompt library and management platform built around DeepSeek, with templates for writing, programming, and creative work, an AI prompt optimizer, and a Chrome extension for faster reuse. The template library can be browsed for free. It will not invent a perfect prompt for every task, but it is much easier than reconstructing a good one from chat history.]]&gt;</description>
      <content:encoded><![CDATA[<p>When a prompt finally works well in DeepSeek, I used to leave it buried in an old chat. The next time I needed the same pattern — a writing rewrite, a coding task, or a creative brief — I would scroll, copy fragments, and slightly change the wording without remembering why the original worked.</p>

<p>The loop I use now looks like this:</p>
<ul><li>Start from a ready-made template that is already close to the task, instead of a blank prompt.</li>
<li>Adjust the constraints in a management workspace so the prompt is reusable, not tied to one conversation.</li>
<li>Run the wording through a prompt optimizer when the output is close but inconsistent.</li>
<li>Keep the saved version somewhere I can open quickly, including from the browser.</li></ul>

<p>I have been using <a href="https://deepseekprompt.top" rel="nofollow">DeepSeek Prompt</a> for this. It is a prompt library and management platform built around DeepSeek, with templates for writing, programming, and creative work, an AI prompt optimizer, and a Chrome extension for faster reuse. The template library can be browsed for free. It will not invent a perfect prompt for every task, but it is much easier than reconstructing a good one from chat history.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/keeping-deepseek-prompts-in-a-library-instead-of-old-chat-threads</guid>
      <pubDate>Fri, 13 Nov 2026 05:06:36 +0000</pubDate>
    </item>
    <item>
      <title>Estimating a body-shape category from one full-body photo</title>
      <link>https://rant.li/pengliangli/estimating-a-body-shape-category-from-one-full-body-photo</link>
      <description>&lt;![CDATA[When I am deciding which cuts of clothing to try first, I usually start with a rough silhouette category — hourglass, pear, apple, rectangle, or inverted triangle. I do not want to take tape measurements just for that first pass, and I also do not want a tool that treats those labels as medical facts.&#xA;&#xA;The workflow I use is straightforward:&#xA;&#xA;Take one clear, front-facing full-body photo in even light, with the outline of the shoulders, waist, and hips visible.&#xA;Upload that photo and read the closest of the five common categories, including any confidence cues.&#xA;Check the written explanation against what I can actually see in the photo — shoulder, waist, and hip balance — rather than accepting a label alone.&#xA;Use the grouped suggestions for tops, bottoms, dresses, and outerwear as a starting list, not a fixed uniform.&#xA;Keep the result as a style-oriented visual estimate. It is not a tailoring measurement and it is not a health assessment.&#xA;&#xA;I have been using AI Body Shape Analyzer for this. It estimates a category from one photo, explains the visible cues, and groups practical styling ideas, including a downloadable advice card. No account or tape measure is required, and photos are processed for one-time analysis rather than saved to the site&#39;s database. It will not replace a fitting or a tailor, but it is useful when I want a structured starting point for outfit planning.]]&gt;</description>
      <content:encoded><![CDATA[<p>When I am deciding which cuts of clothing to try first, I usually start with a rough silhouette category — hourglass, pear, apple, rectangle, or inverted triangle. I do not want to take tape measurements just for that first pass, and I also do not want a tool that treats those labels as medical facts.</p>

<p>The workflow I use is straightforward:</p>
<ul><li>Take one clear, front-facing full-body photo in even light, with the outline of the shoulders, waist, and hips visible.</li>
<li>Upload that photo and read the closest of the five common categories, including any confidence cues.</li>
<li>Check the written explanation against what I can actually see in the photo — shoulder, waist, and hip balance — rather than accepting a label alone.</li>
<li>Use the grouped suggestions for tops, bottoms, dresses, and outerwear as a starting list, not a fixed uniform.</li>
<li>Keep the result as a style-oriented visual estimate. It is not a tailoring measurement and it is not a health assessment.</li></ul>

<p>I have been using <a href="https://aibodyshapeanalyzer.com" rel="nofollow">AI Body Shape Analyzer</a> for this. It estimates a category from one photo, explains the visible cues, and groups practical styling ideas, including a downloadable advice card. No account or tape measure is required, and photos are processed for one-time analysis rather than saved to the site&#39;s database. It will not replace a fitting or a tailor, but it is useful when I want a structured starting point for outfit planning.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/estimating-a-body-shape-category-from-one-full-body-photo</guid>
      <pubDate>Fri, 13 Nov 2026 05:06:21 +0000</pubDate>
    </item>
    <item>
      <title>Turning an idea into an anime scene on an infinite canvas</title>
      <link>https://rant.li/pengliangli/turning-an-idea-into-an-anime-scene-on-an-infinite-canvas</link>
      <description>&lt;![CDATA[I often start with only a loose idea — a character at a train-platform vending machine, or a quiet rooftop after rain — and try to turn it into a readable anime scene. Working on an infinite canvas in the browser helps: I can drop a rough composition, shove the camera left or right, and keep alternate panels beside the one I&#39;m actually committing to, instead of flipping between files.&#xA;&#xA;Lately I&#39;ve been using Anime AI Studio for this. I&#39;ll sketch the beat on the canvas, generate a still to see whether the lighting and staging hold, then adjust the posing or the crop before I write the next shot. It keeps the drafting loop on one page, which is what I want when a scene is still finding its shape.]]&gt;</description>
      <content:encoded><![CDATA[<p>I often start with only a loose idea — a character at a train-platform vending machine, or a quiet rooftop after rain — and try to turn it into a readable anime scene. Working on an infinite canvas in the browser helps: I can drop a rough composition, shove the camera left or right, and keep alternate panels beside the one I&#39;m actually committing to, instead of flipping between files.</p>

<p>Lately I&#39;ve been using <a href="https://animeaistudio.com" rel="nofollow">Anime AI Studio</a> for this. I&#39;ll sketch the beat on the canvas, generate a still to see whether the lighting and staging hold, then adjust the posing or the crop before I write the next shot. It keeps the drafting loop on one page, which is what I want when a scene is still finding its shape.</p>
]]></content:encoded>
      <guid>https://rant.li/pengliangli/turning-an-idea-into-an-anime-scene-on-an-infinite-canvas</guid>
      <pubDate>Fri, 13 Nov 2026 02:59:35 +0000</pubDate>
    </item>
  </channel>
</rss>