<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:webfeeds="http://webfeeds.org/rss/1.0"
	>
<channel>
	<title>
	Comments for IK3D	</title>
	<atom:link href="https://www.ik3d.fr/comments/feed/" rel="self" type="application/rss+xml" />
	<link>https://ik3d.fr/</link>
	<description>Art Director / B3D</description>
	<lastBuildDate>Wed, 09 Sep 2026 14:19:28 +0000</lastBuildDate>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	
	<item>
		<title>
		Comment on Rodin Gen-2.5 — The First 10-Million-Polygon AI 3D Model, and Its Best Trick Isn&#8217;t the Geometry by Muse		</title>
		<link>https://www.ik3d.fr/rodin-gen-2-5-the-first-10-million-polygon-ai-3d-model-and-its-best-trick-isnt-the-geometry/#comment-139842</link>

		<dc:creator><![CDATA[Muse]]></dc:creator>
		<pubDate>Wed, 09 Sep 2026 14:19:28 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47640#comment-139842</guid>

					<description><![CDATA[Rodin Gen-2.5 is impressive, but for video, I keep coming back to muse photo to video. It turns images into clips with native audio so effortlessly—plus the 4K output without a watermark? That’s a game changer for my workflow. &lt;a href=&quot;https://muse-video.co
&quot;&gt;muse photo to video&lt;/a&gt;]]></description>
			<content:encoded><![CDATA[<p>Rodin Gen-2.5 is impressive, but for video, I keep coming back to muse photo to video. It turns images into clips with native audio so effortlessly—plus the 4K output without a watermark? That’s a game changer for my workflow. <a href="https://muse-video.co
">muse photo to video</a></p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Seedance 2.5 — ByteDance Says the Stitching Era Is Over: One Unbroken 30-Second AI Video Shot, 50 References, and Edits That Don&#8217;t Nuke the Whole Clip by frame to video		</title>
		<link>https://www.ik3d.fr/seedance-2-5-bytedance-says-the-stitching-era-is-over-one-unbroken-30-second-ai-video-shot-50-references-and-edits-that-dont-nuke-the-whole-clip/#comment-139831</link>

		<dc:creator><![CDATA[frame to video]]></dc:creator>
		<pubDate>Sat, 05 Sep 2026 15:48:05 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47863#comment-139831</guid>

					<description><![CDATA[Practical use case: editing a 30-second product hero video where the client wants one label changed. Today that means regenerating everything and hoping nothing else moves. Seedance 2.5’s region-level editing plus multi-reference control would make it a real edit session instead of a gamble — that’s the part that excites me most. The article frames the shift well without overselling. I’ve also been collecting related AI video workflow notes and breakdowns on https://framov.com 
 and this analysis fits nicely there. Keen to see independent tests.]]></description>
			<content:encoded><![CDATA[<p>Practical use case: editing a 30-second product hero video where the client wants one label changed. Today that means regenerating everything and hoping nothing else moves. Seedance 2.5’s region-level editing plus multi-reference control would make it a real edit session instead of a gamble — that’s the part that excites me most. The article frames the shift well without overselling. I’ve also been collecting related AI video workflow notes and breakdowns on <a href="https://framov.com" rel="nofollow ugc">https://framov.com</a><br />
 and this analysis fits nicely there. Keen to see independent tests.</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Seedance 2.5 — ByteDance Says the Stitching Era Is Over: One Unbroken 30-Second AI Video Shot, 50 References, and Edits That Don&#8217;t Nuke the Whole Clip by Hibt		</title>
		<link>https://www.ik3d.fr/seedance-2-5-bytedance-says-the-stitching-era-is-over-one-unbroken-30-second-ai-video-shot-50-references-and-edits-that-dont-nuke-the-whole-clip/#comment-139827</link>

		<dc:creator><![CDATA[Hibt]]></dc:creator>
		<pubDate>Thu, 03 Sep 2026 12:42:28 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47863#comment-139827</guid>

					<description><![CDATA[Nếu bạn đã quen với biến động của &lt;a href=&quot;https://hibt.com/vi/symbolPrediction/detail/BTC&quot; rel=&quot;nofollow ugc&quot;&gt;BTC – Bitcoin&lt;/a&gt; (https://hibt.com/vi/symbolPrediction/detail/BTC) , hãy nhớ rằng MOONSHOTUSDT có mức độ rủi ro cao hơn nhiều do thiếu thị trường tham chiếu công khai]]></description>
			<content:encoded><![CDATA[<p>Nếu bạn đã quen với biến động của <a href="https://hibt.com/vi/symbolPrediction/detail/BTC" rel="nofollow ugc">BTC – Bitcoin</a> (<a href="https://hibt.com/vi/symbolPrediction/detail/BTC" rel="nofollow ugc">https://hibt.com/vi/symbolPrediction/detail/BTC</a>) , hãy nhớ rằng MOONSHOTUSDT có mức độ rủi ro cao hơn nhiều do thiếu thị trường tham chiếu công khai</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Seedance 2.5 — ByteDance Says the Stitching Era Is Over: One Unbroken 30-Second AI Video Shot, 50 References, and Edits That Don&#8217;t Nuke the Whole Clip by frame to video		</title>
		<link>https://www.ik3d.fr/seedance-2-5-bytedance-says-the-stitching-era-is-over-one-unbroken-30-second-ai-video-shot-50-references-and-edits-that-dont-nuke-the-whole-clip/#comment-139822</link>

		<dc:creator><![CDATA[frame to video]]></dc:creator>
		<pubDate>Wed, 02 Sep 2026 16:44:43 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47863#comment-139822</guid>

					<description><![CDATA[The claim that actually made me pause isn’t the 30-second single pass—it’s the consistency-preserving local edit. Fix one prop without regenerating the entire clip? That changes the workflow more than any duration bump. No more hoping the next generation keeps the other 28 seconds intact. I’ve been collecting practical notes on shot-level AI iteration at framov.com, and your breakdown of the caveats is a valuable reality check. Closed beta, no third-party benchmarks—I’ll wait before reshuffling my pipeline. Still, this feels like a meaningful direction.

https://framov.com]]></description>
			<content:encoded><![CDATA[<p>The claim that actually made me pause isn’t the 30-second single pass—it’s the consistency-preserving local edit. Fix one prop without regenerating the entire clip? That changes the workflow more than any duration bump. No more hoping the next generation keeps the other 28 seconds intact. I’ve been collecting practical notes on shot-level AI iteration at framov.com, and your breakdown of the caveats is a valuable reality check. Closed beta, no third-party benchmarks—I’ll wait before reshuffling my pipeline. Still, this feels like a meaningful direction.</p>
<p><a href="https://framov.com" rel="nofollow ugc">https://framov.com</a></p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Seedance 2.5 — ByteDance Says the Stitching Era Is Over: One Unbroken 30-Second AI Video Shot, 50 References, and Edits That Don&#8217;t Nuke the Whole Clip by frame to video		</title>
		<link>https://www.ik3d.fr/seedance-2-5-bytedance-says-the-stitching-era-is-over-one-unbroken-30-second-ai-video-shot-50-references-and-edits-that-dont-nuke-the-whole-clip/#comment-139821</link>

		<dc:creator><![CDATA[frame to video]]></dc:creator>
		<pubDate>Wed, 02 Sep 2026 16:40:37 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47863#comment-139821</guid>

					<description><![CDATA[Great breakdown. I agree that native 30-second generation feels like the real ceiling shift, but the consistency-preserving local edit is what makes this practical for directors — fixing one element without regenerating the whole timeline is the feature that actually saves production time. As you note, third-party verification is still pending, so I’ll hold some excitement back. For anyone tracking these AI video developments and looking for workflow comparisons, I recently found some useful references at https://framov.com. Thanks for the thorough analysis.]]></description>
			<content:encoded><![CDATA[<p>Great breakdown. I agree that native 30-second generation feels like the real ceiling shift, but the consistency-preserving local edit is what makes this practical for directors — fixing one element without regenerating the whole timeline is the feature that actually saves production time. As you note, third-party verification is still pending, so I’ll hold some excitement back. For anyone tracking these AI video developments and looking for workflow comparisons, I recently found some useful references at <a href="https://framov.com" rel="nofollow ugc">https://framov.com</a>. Thanks for the thorough analysis.</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on VARCO 3D — The Studio Behind Lineage Just Shipped Text-to-Animated-3D in Your Browser, for 35 Cents an Asset by Guest		</title>
		<link>https://www.ik3d.fr/varco-3d-the-studio-behind-lineage-just-shipped-text-to-animated-3d-in-your-browser-for-35-cents-an-asset/#comment-139806</link>

		<dc:creator><![CDATA[Guest]]></dc:creator>
		<pubDate>Tue, 01 Sep 2026 04:23:32 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47600#comment-139806</guid>

					<description><![CDATA[If you&#039;re into planning outdoor spaces too, LandscapeAI is worth a peek—you upload a photo of your yard and it shows different styles. Try it here: &lt;a href=&quot;https://landscapedesignai.net/
&quot;&gt;ai landscape design&lt;/a&gt;]]></description>
			<content:encoded><![CDATA[<p>If you&#8217;re into planning outdoor spaces too, LandscapeAI is worth a peek—you upload a photo of your yard and it shows different styles. Try it here: <a href="https://landscapedesignai.net/
">ai landscape design</a></p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Revit 2027 Just Put an AI Agent Inside the World&#8217;s BIM Standard — And Wired It for MCP, So Claude Can Read Your Live Building Model by dliyatul awliya		</title>
		<link>https://www.ik3d.fr/revit-2027-just-put-an-ai-agent-inside-the-worlds-bim-standard-and-wired-it-for-mcp-so-claude-can-read-your-live-building-model/#comment-139805</link>

		<dc:creator><![CDATA[dliyatul awliya]]></dc:creator>
		<pubDate>Tue, 01 Sep 2026 02:25:25 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47996#comment-139805</guid>

					<description><![CDATA[The integration of AI agents and MCP into Revit 2027 is a fascinating step for the AEC industry. Being able to interact with live BIM data through natural-language prompts could make complex workflows much faster and more efficient.]]></description>
			<content:encoded><![CDATA[<p>The integration of AI agents and MCP into Revit 2027 is a fascinating step for the AEC industry. Being able to interact with live BIM data through natural-language prompts could make complex workflows much faster and more efficient.</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Seedance 2.5 — ByteDance Says the Stitching Era Is Over: One Unbroken 30-Second AI Video Shot, 50 References, and Edits That Don&#8217;t Nuke the Whole Clip by frame to video		</title>
		<link>https://www.ik3d.fr/seedance-2-5-bytedance-says-the-stitching-era-is-over-one-unbroken-30-second-ai-video-shot-50-references-and-edits-that-dont-nuke-the-whole-clip/#comment-139800</link>

		<dc:creator><![CDATA[frame to video]]></dc:creator>
		<pubDate>Mon, 31 Aug 2026 16:34:50 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47863#comment-139800</guid>

					<description><![CDATA[Great breakdown of Seedance 2.5’s real implications. The point about region-level editing being the bigger leap than raw duration resonates — preserving 28 good seconds instead of rerolling the whole shot is what makes AI video feel like actual art direction. The 50-reference input also sounds like a practical way to keep visual identity locked across a full scene. For anyone digging into these workflow shifts, I recently came across a companion resource covering similar AI production concepts at https://framov.com — worth a look if you’re testing whether these tools fit into a real pipeline.]]></description>
			<content:encoded><![CDATA[<p>Great breakdown of Seedance 2.5’s real implications. The point about region-level editing being the bigger leap than raw duration resonates — preserving 28 good seconds instead of rerolling the whole shot is what makes AI video feel like actual art direction. The 50-reference input also sounds like a practical way to keep visual identity locked across a full scene. For anyone digging into these workflow shifts, I recently came across a companion resource covering similar AI production concepts at <a href="https://framov.com" rel="nofollow ugc">https://framov.com</a> — worth a look if you’re testing whether these tools fit into a real pipeline.</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on Seedance 2.5 — ByteDance Says the Stitching Era Is Over: One Unbroken 30-Second AI Video Shot, 50 References, and Edits That Don&#8217;t Nuke the Whole Clip by frame to video		</title>
		<link>https://www.ik3d.fr/seedance-2-5-bytedance-says-the-stitching-era-is-over-one-unbroken-30-second-ai-video-shot-50-references-and-edits-that-dont-nuke-the-whole-clip/#comment-139799</link>

		<dc:creator><![CDATA[frame to video]]></dc:creator>
		<pubDate>Mon, 31 Aug 2026 16:30:20 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47863#comment-139799</guid>

					<description><![CDATA[Thirty seconds in one take, 50 reference inputs, and local edits that don’t nuke the whole clip — is that enough to call the stitching era over? The insight that hits me is the consistency-preserving edit: that turns AI video from a slot machine into a directable timeline. I’ve been exploring similar ideas around keeping character and product references locked across motion, and I found some practical workflow notes and tool comparisons on &lt;a href=&quot;https://framov.com/
&quot;&gt;framov.com&lt;/a&gt;. Definitely worth a read if this workflow resonates. Appreciate the clear breakdown.]]></description>
			<content:encoded><![CDATA[<p>Thirty seconds in one take, 50 reference inputs, and local edits that don’t nuke the whole clip — is that enough to call the stitching era over? The insight that hits me is the consistency-preserving edit: that turns AI video from a slot machine into a directable timeline. I’ve been exploring similar ideas around keeping character and product references locked across motion, and I found some practical workflow notes and tool comparisons on <a href="https://framov.com/
">framov.com</a>. Definitely worth a read if this workflow resonates. Appreciate the clear breakdown.</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		Comment on NeRF Just Won Computing&#8217;s Top Early-Career Prize — and Almost Every 3D Tool the Lab Covers Traces Back to These Two Researchers by dliyatul awliya		</title>
		<link>https://www.ik3d.fr/nerf-just-won-computings-top-early-career-prize-and-almost-every-3d-tool-the-lab-covers-traces-back-to-these-two-researchers/#comment-139798</link>

		<dc:creator><![CDATA[dliyatul awliya]]></dc:creator>
		<pubDate>Mon, 31 Aug 2026 14:44:12 +0000</pubDate>
		<guid isPermaLink="false">https://ik3d.fr/?p=47685#comment-139798</guid>

					<description><![CDATA[A fascinating look at how NeRF laid the groundwork for today’s 3D AI technologies. The connection to Gaussian splatting and generative 3D makes this a really insightful read.]]></description>
			<content:encoded><![CDATA[<p>A fascinating look at how NeRF laid the groundwork for today’s 3D AI technologies. The connection to Gaussian splatting and generative 3D makes this a really insightful read.</p>
]]></content:encoded>
		
			</item>
	</channel>
</rss>
