<?xml version="1.0" encoding="UTF-8"?>
    <rss
      version="2.0"
      xmlns:dc="http://purl.org/dc/elements/1.1/"
      xmlns:content="http://purl.org/rss/1.0/modules/content/"
      xmlns:atom="http://www.w3.org/2005/Atom"
    >
      <channel>
        <title><![CDATA[Runware Docs Changelog]]></title>
        <description><![CDATA[The latest updates and improvements to the Runware platform.]]></description>
        <link>https://runware.ai/docs/changelog</link>
        <generator>Astro</generator>
        <lastBuildDate>Thu, 19 Nov 2026 23:18:18 GMT</lastBuildDate>
        <atom:link
          href="https://runware.ai/docs/changelog/rss.xml"
          rel="self"
          type="application/rss+xml"
        />
        <pubDate>Thu, 19 Nov 2026 00:00:00 GMT</pubDate>
        <copyright><![CDATA[Copyright (C) 2026 Runware]]></copyright>
        <language>en</language>
        <ttl>1440</ttl>
        
      <item>
        <title><![CDATA[Ideogram 4.6 and Precise Edit]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-19-ideogram-4-6</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Thu, 19 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Ideogram 4.6 and Ideogram 4.6 Precise Edit are now available on Runware for image generation and instruction-guided editing at up to 4K.</p>
<section id="ideogram-46"><h3 id="ideogram-46">Ideogram 4.6</h3><ul>
<li>Use <code>ideogram:4.6@0</code> to generate images from text at any size from 256 to 4096 pixels per side, in multiples of 32 pixels with an aspect ratio between 1:6 and 6:1.</li>
<li>Edit a source image with up to four additional reference images, each up to 25 MB.</li>
<li>Set <code>settings.background</code> to <code>transparent</code> for output with an alpha channel, which requires <code>outputFormat</code> <code>PNG</code> or <code>WEBP</code>, or to <code>opaque</code> for a solid background.</li>
<li><code>settings.quality</code> takes <code>very_low</code>, <code>low</code>, <code>medium</code> or <code>high</code>, with sizes up to 4,194,304 pixels at <code>very_low</code>. <code>settings.magicPrompt</code> controls prompt enhancement on text-to-image requests.</li>
</ul></section>
<section id="ideogram-46-precise-edit"><h3 id="ideogram-46-precise-edit">Ideogram 4.6 Precise Edit</h3><ul>
<li>Use <code>ideogram:4.6@precise-edit</code> to apply a text instruction to a required <code>inputs.seedImage</code> of up to 8192×8192 pixels and 50 MB. The output keeps the source dimensions.</li>
<li>Use <code>inputs.maskImage</code> to constrain changes and up to four reference images to guide the edit, or three with a mask.</li>
<li>Set <code>settings.editRegion</code> to a box of <code>[x1, y1, x2, y2]</code> pixel coordinates to edit part of a 4K or 8K image at full resolution, or to <code>none</code> to send the whole image. The box must lie inside the image, with each side at least 256 pixels and an aspect ratio between 1:6 and 6:1.</li>
</ul></section>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Kling VIDEO 4.0 4K and HDR]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-19-kling-video-4-0-4k-hdr</guid>
        <category><![CDATA[feature]]></category>
        <pubDate>Thu, 19 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Kling VIDEO 4.0 adds native 4K output to its 720p and 1080p tiers, and 10-bit HDR output.</p>
<ul>
<li>Set <code>resolution</code> to <code>4k</code>, or set <code>width</code> and <code>height</code> to one of four 4K sizes: 3840×2160 (16:9), 2880×2880 (1:1), 2160×3840 (9:16) or 4396×1884 (21:9).</li>
<li>With any image or video input, set the tier with <code>resolution</code> and the inputs set the aspect ratio.</li>
<li>Set <code>settings.dynamicRange</code> to <code>hdr</code> for 10-bit HEVC in BT.2020 with HLG, at 1080p or 4K with <code>outputFormat</code> <code>MP4</code>. The default <code>sdr</code> returns 8-bit H.264.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Kling IMAGE 4.0]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-17-kling-image-4-0</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Tue, 17 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Kling IMAGE 4.0 is now available on Runware as a single model for text-to-image, reference-guided generation, image series and long-form page design.</p>
<ul>
<li>Use <code>klingai:kling-image@4.0</code> to generate at 2K, 4K or 8K, with <code>resolution</code> alone so the model picks the aspect ratio, or with a <code>width</code> and <code>height</code> pair in eight aspect ratios from 21:9 to 9:16.</li>
<li>Accepts up to 10 reference images and image-based elements combined, cited in the prompt as <code>&lt;&lt;&lt;image_1&gt;&gt;&gt;</code> and <code>&lt;&lt;&lt;element_1&gt;&gt;&gt;</code>, with prompts of up to 50,000 characters.</li>
<li>Generates a series of 2 to 10 related images that keep subjects and style consistent with <code>settings.seriesLength</code>, or lets the model choose the count with <code>auto</code>.</li>
<li>Designs a website, or an e-commerce product page from 1 to 3 product images, as one tall image at 1K or 2K with <code>settings.pageType</code>. <code>settings.pageLength</code> sets how many sections the product page has.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[GPT-6.2 Astra and Sol]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-17-openai-gpt-6-2</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Tue, 17 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>GPT-6.2 Astra and GPT-6.2 Sol are now available through Runware&#39;s <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI Chat Completions</a> and <a href="/docs/tools/compatible-apis#openai-responses">OpenAI Responses</a> endpoints.</p>
<ul>
<li>Use <code>openai:gpt@6.2-astra</code> or <code>openai:gpt@6.2-sol</code> as the model value on <code>/v1/chat/completions</code> or <code>/v1/responses</code>.</li>
<li>Both support text and image inputs, structured outputs, function calling, and tool use.</li>
<li>Both provide a 1,050,000-token context window with up to 128,000 output tokens and configurable reasoning from low through max.</li>
<li>GPT-6.2 Astra targets complex professional work: software engineering, scientific research, documents, and computer use.</li>
<li>GPT-6.2 Sol is the efficient tier for everyday coding, writing, and high-volume agent workflows.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[HeyGen Video Translate 1.0]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-16-heygen-video-translate-1-0</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>HeyGen Video Translate 1.0 is now available on Runware for translating the speech in a video into another language, in the speaker&#39;s own voice and with matching lip sync.</p>
<ul>
<li>Use <code>heygen:video-translate@1.0</code> with the source video in <code>inputs.video</code> and the target language in <code>speech.language</code>, one of 190 language codes. The source language is detected, or set with <code>inputs.video.language</code>.</li>
<li><code>settings.mode</code> selects <code>speed</code> for faster turnaround or <code>precision</code> for higher lip sync quality.</li>
<li><code>settings.translateAudioOnly</code> dubs the speech and keeps the original picture, and <code>settings.startTime</code> with <code>settings.endTime</code> translates only part of the video.</li>
<li>Accepts a custom dubbing track in <code>inputs.audio</code> and an SRT file in <code>inputs.subtitles</code>, as the source transcript or as the translated lines.</li>
<li><code>settings.speakerCount</code> separates up to 10 speakers, <code>settings.removeBackgroundMusic</code> drops the music track and <code>settings.enhanceSpeech</code> cleans up the speech.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Qwen4-Max]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-16-qwen4-max</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Qwen4-Max is now available through Runware&#39;s <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI Chat Completions</a> and <a href="/docs/tools/compatible-apis#openai-responses">OpenAI Responses</a> endpoints.</p>
<ul>
<li>Use <code>alibaba:qwen@4-max</code> as the model value on <code>/v1/chat/completions</code> or <code>/v1/responses</code>.</li>
<li>Supports text, image, and video inputs, tool calling, and structured outputs for coding, office work, front-end design, and agent workflows.</li>
<li>Provides a 1,000,000-token context window with up to 131,072 output tokens and configurable reasoning effort.</li>
<li>Builds on Qwen3.8-Max with stronger world knowledge, perception, and professional tool use.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Seedance 2.5 4K]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-16-seedance-2-5-4k</guid>
        <category><![CDATA[feature]]></category>
        <pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Seedance 2.5 adds 4K output to its 480p, 720p and 1080p tiers.</p>
<ul>
<li>Set <code>resolution</code> to <code>4k</code>, or set <code>width</code> and <code>height</code> to one of six 4K sizes: 3840×2160 (16:9), 3326×2494 (4:3), 2880×2880 (1:1), 2494×3326 (3:4), 2160×3840 (9:16) or 4398×1886 (21:9).</li>
<li>With a first or last frame, or a source video to edit or extend, set the tier with <code>resolution</code> and the input sets the aspect ratio.</li>
<li>To render an approved draft (<code>inputs.draftId</code>) in 4K, send <code>resolution</code> set to <code>4k</code> or a 4K <code>width</code> and <code>height</code>. Without either, the draft renders at 1080p. Draft mode itself (<code>settings.draft</code>) still generates at 480p.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Seedance 2.5 Flash]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-15-seedance-2-5-flash</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Seedance 2.5 Flash is a lower-cost variant of ByteDance&#39;s Seedance 2.5 video model, with the same multimodal generation, editing and extension workflows at 480p and 720p.</p>
<ul>
<li>Use <code>bytedance:seedance@2.5-flash</code> to generate 4 to 30 second videos with synchronized audio from a prompt of up to 50,000 characters, or set <code>duration</code> to <code>auto</code> to let the model choose the length.</li>
<li>Starts from a first frame or a first and last frame, or takes up to 10 reference images, 5 reference videos and 5 reference audio clips, cited in the prompt as <code>@Image1</code>, <code>@Video1</code> and <code>@Audio1</code>.</li>
<li>Edits or extends a source clip in <code>inputs.video</code>, with <code>settings.operation</code> set to <code>edit</code> or <code>extend</code>, or left to the prompt.</li>
<li>Sets the size with <code>resolution</code> (<code>480p</code> or <code>720p</code>), or with a <code>width</code> and <code>height</code> pair in six aspect ratios from 21:9 to 9:16 at each tier.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Gemini Omni Flash 1.1 Deprecation]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-13-gemini-omni-flash-1-1-deprecated</guid>
        <category><![CDATA[deprecation]]></category>
        <pubDate>Fri, 13 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Gemini Omni Flash 1.1 is deprecated and will be removed from Runware on <strong>December 3, 2026</strong>. After that date, requests using <code>google:gemini@omni-flash-1.1</code> will return an error.</p>
<section id="migration"><h3 id="migration">Migration</h3><ul>
<li>Migrate Gemini Omni Flash 1.1 (<code>google:gemini@omni-flash-1.1</code>) to <a href="/docs/models/google-gemini-omni-flash-1-2">Gemini Omni Flash 1.2</a> (<code>google:gemini@omni-flash-1.2</code>).</li>
<li>Gemini Omni Flash 1.2 takes the same inputs, raises the reference image limit from 7 to 10 and the reference image file size from 5 MB to 20 MB, and raises the maximum <code>duration</code> from 10 to 15 seconds for text-to-video and single-frame image-to-video.</li>
<li>An input video is limited to 32 MB instead of 48 MB, and to 27 seconds instead of 30 when <code>duration</code> is set.</li>
</ul></section>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Gemini 4 Hydrogen]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-13-google-gemini-4-hydrogen</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Fri, 13 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Gemini 4 Hydrogen is now available through Runware&#39;s <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI-compatible layer</a>.</p>
<ul>
<li>Use <code>google:gemini@4-hydrogen</code> as the model value in OpenAI-compatible requests.</li>
<li>Supports text, image, video, audio, and PDF inputs for multimodal understanding and text-generation workflows.</li>
<li>Provides a 1,048,576-token context window and configurable thinking effort.</li>
<li>Built for everyday coding, agent workflows, and document-heavy knowledge work, including worker and verifier agents in multi-model workflows.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[GPT-5.4 Nano Deprecation]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-13-gpt-5-4-nano-deprecated</guid>
        <category><![CDATA[deprecation]]></category>
        <pubDate>Fri, 13 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>GPT-5.4 Nano is deprecated and will be removed from Runware on <strong>April 1, 2027</strong>. After that date, requests using <code>openai:gpt@5.4-nano</code> will return an error.</p>
<section id="migration"><h3 id="migration">Migration</h3><ul>
<li>Migrate GPT-5.4 Nano (<code>openai:gpt@5.4-nano</code>) to GPT-6 Luna (<code>openai:gpt@6-luna</code>) on the <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI Chat Completions</a> or <a href="/docs/tools/compatible-apis#openai-responses">OpenAI Responses</a> endpoint. GPT-6 Luna is not available on the native <code>textInference</code> task.</li>
<li>GPT-6 Luna takes the same text and image inputs, keeps the 128,000-token output limit and raises the context window from 400,000 to 1,050,000 tokens.</li>
</ul></section>]]></description>
      </item>
    
      <item>
        <title><![CDATA[GPT Image 2 Deprecation]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-13-gpt-image-2-deprecated</guid>
        <category><![CDATA[deprecation]]></category>
        <pubDate>Fri, 13 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>GPT Image 2 is deprecated and will be removed from Runware on <strong>February 10, 2027</strong>. After that date, requests using <code>openai:gpt-image@2</code> will return an error.</p>
<section id="migration"><h3 id="migration">Migration</h3><ul>
<li>Migrate GPT Image 2 (<code>openai:gpt-image@2</code>) to <a href="/docs/models/openai-gpt-image-2-6-sunburst">GPT-Image-2.6 Sunburst</a> (<code>openai:gpt-image@2.6-sunburst</code>) for detailed compositions and precise edits, or to <a href="/docs/models/openai-gpt-image-2-6-flare">GPT-Image-2.6 Flare</a> (<code>openai:gpt-image@2.6-flare</code>) for faster iteration.</li>
<li>Both take the same reference and mask images, and add markup images, SVG and GIF output, transparent backgrounds and <code>thinkingLevel</code>.</li>
</ul></section>]]></description>
      </item>
    
      <item>
        <title><![CDATA[MiniMax M2.7 and M2.7 Highspeed Deprecation]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-12-minimax-m2-7-deprecated</guid>
        <category><![CDATA[deprecation]]></category>
        <pubDate>Thu, 12 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>MiniMax M2.7 and MiniMax M2.7 Highspeed are deprecated and will be removed from Runware on <strong>November 18, 2026</strong>. After that date, requests using <code>minimax:m2.7@0</code> or <code>minimax:m2.7@highspeed</code> will return an error.</p>
<section id="migration"><h3 id="migration">Migration</h3><ul>
<li>Migrate MiniMax M2.7 (<code>minimax:m2.7@0</code>) and MiniMax M2.7 Highspeed (<code>minimax:m2.7@highspeed</code>) to <a href="/docs/models/minimax-m3">MiniMax M3</a> (<code>minimax:m3@0</code>).</li>
<li>MiniMax M3 adds image and video input, and raises the context window from 204,800 to 1M tokens and the <code>maxTokens</code> limit from 131,072 to 512,000.</li>
</ul></section>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Seedream 5.5]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-12-seedream-5_5</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Thu, 12 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Seedream 5.5 is ByteDance&#39;s image generation and editing model, with transparent output and decomposition of an existing image into editable layers.</p>
<ul>
<li>Use <code>bytedance:seedream@5.5</code> to generate from a prompt of up to 10,000 characters, or to edit and combine up to 10 reference images while keeping unedited areas stable across multi-turn edits.</li>
<li>Sets the size with <code>resolution</code> (<code>1K</code>, <code>1.5K</code> or <code>2K</code>), in the shape of the references or the one the prompt describes, or with an exact <code>width</code> and <code>height</code> of 921,600 to 4,624,220 pixels in total and aspect ratios from 1:16 to 16:1. With references and no size, it generates at <code>2K</code>.</li>
<li><code>settings.background</code> set to <code>transparent</code> returns a transparent PNG or WebP, from a prompt alone or from references.</li>
<li><code>settings.layerDecomposition</code> splits one reference image into a base image and up to 99 transparent PNG layers, each with its position, stacking order, name and description. <code>settings.layerDecompositionMaxImages</code> caps the total at 2 to 100 images, base image included, and an optional prompt guides the split.</li>
<li><code>settings.layerDecompositionTextRecognition</code> adds the text of each text layer line by line, with its position, color, closest font and whether it is bold, italic, underlined, struck through, shadowed or translucent.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Claude Fable 5.2]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-10-claude-fable-5_2</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Tue, 10 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Claude Fable 5.2 is now available through Runware&#39;s <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI-compatible layer</a> and the <a href="/docs/tools/compatible-apis#anthropic-messages">Anthropic Messages endpoint</a>.</p>
<ul>
<li>Use <code>anthropic:claude@fable-5.2</code> as the model value on <code>/v1/chat/completions</code> or <code>/v1/messages</code>.</li>
<li>Supports text and image inputs, tool calling, and configurable reasoning effort.</li>
<li>Provides a 1M-token context window with up to 128,000 output tokens.</li>
<li>Improves on Claude Fable 5.1 in coding, reasoning, vision, and computer use, and generates output faster.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Gemini 4 Carbon]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-10-google-gemini-4-carbon</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Tue, 10 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Gemini 4 Carbon is now available through Runware&#39;s <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI-compatible layer</a>.</p>
<ul>
<li>Use <code>google:gemini@4-carbon</code> as the model value in OpenAI-compatible requests.</li>
<li>Supports text, image, video, audio, and PDF inputs for multimodal understanding and text-generation workflows.</li>
<li>Provides a 1,048,576-token context window with up to 65,536 output tokens and configurable thinking effort.</li>
<li>Built for the hardest software engineering, enterprise knowledge work, and scientific reasoning tasks, spending more tokens than Gemini 4 Argon to reason and verify its work.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[MiniMax H3 Image and H3 Image Highspeed]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-10-minimax-h3-image</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Tue, 10 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>MiniMax H3 Image and MiniMax H3 Image Highspeed are now available for image generation and multi-reference editing, at up to 20 megapixels.</p>
<ul>
<li><strong>H3 Image</strong> (<code>minimax:h3@image</code>) is MiniMax&#39;s open-weights image generation and editing model.</li>
<li><strong>H3 Image Highspeed</strong> (<code>minimax:h3@image-highspeed</code>) is the faster variant, with the same parameters and limits.</li>
<li>Both generate from a prompt of up to 32,000 characters, or edit and combine up to 16 reference images, each 128 to 8,192 pixels per side with an aspect ratio of up to 4:1.</li>
<li>Both set the size with <code>resolution</code> (<code>1K</code>, <code>2K</code> or <code>4K</code>), in the shape of the references or one the model picks from the prompt, or with an exact <code>width</code> and <code>height</code> of 655,360 to 20,000,000 pixels in total, 468 to 7,745 pixels per side and aspect ratios from 1:3 to 3:1. With no size, they generate at <code>4K</code>.</li>
<li><code>settings.promptOptimizer</code>, on by default, expands the prompt before generating. Turning it off uses the prompt as written and needs <code>width</code> and <code>height</code>, and is not available with reference images.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Gemini Omni Flash 1.2]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-05-gemini-omni-flash-1_2</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Gemini Omni Flash 1.2 is Google&#39;s multimodal video model for generating, editing and extending clips with native audio, with text-to-video and image-to-video clips up to 15 seconds.</p>
<ul>
<li>Use <code>google:gemini@omni-flash-1.2</code> to generate 3 to 15 second videos from text or a single starting frame, at 360p, 720p, 1080p or 4K in 16:9 and 9:16.</li>
<li>Interpolates between a first and a last frame, and generates from up to 10 reference images and 3 reference videos in the same request, in clips of 3 to 10 seconds.</li>
<li>Edits a source video of 3 to 15 seconds from a prompt when <code>duration</code> is left out, and takes reference images alongside it to bring in new elements.</li>
<li>Extends a source video of up to 27 seconds by 3 to 10 seconds when <code>duration</code> is set.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[GPT-Image-2.6 Flare and Sunburst]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-05-openai-gpt-image-2_6</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst are now available for image generation and editing, with SVG and animated GIF output.</p>
<ul>
<li><strong>Flare</strong> (<code>openai:gpt-image@2.6-flare</code>) is built for fast iteration in interactive workflows.</li>
<li><strong>Sunburst</strong> (<code>openai:gpt-image@2.6-sunburst</code>) is built for detailed compositions and precise editing.</li>
<li>Both support text-to-image generation, prompt-guided editing, masks, and up to 16 reference images, with custom dimensions up to 3,840 pixels per side.</li>
<li><code>inputs.markupImage</code> takes a transparent PNG with marks drawn over the first reference image, like a circle around the object to change, as an alternative to <code>inputs.maskImage</code>.</li>
<li><code>outputFormat</code> adds <code>SVG</code> and <code>GIF</code> for text-to-image. A GIF has 2 to 16 frames (<code>settings.gif.frameCount</code>) of at least 0.02 seconds each (<code>settings.gif.frameDuration</code>), up to 960 pixels per side.</li>
<li><code>settings.svg.model</code> picks the model that writes the SVG, GPT-5.4 Mini by default, or GPT-5.4, GPT-5.5, GPT-5.6 Terra, GPT-6 Luna or GPT-6 Astra, and <code>settings.svg.thinkingLevel</code> sets its reasoning level.</li>
<li><code>settings.thinkingLevel</code> sets how much the model plans before rendering, <code>low</code>, <code>medium</code> or <code>high</code> (the default), independently of <code>settings.quality</code>, which goes from <code>low</code> to <code>max</code>.</li>
<li>Transparent backgrounds are available with PNG or WebP output.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[P-Video-2-Pro-Replace]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-05-prunaai-p-video-2-pro-replace</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>P-Video-2-Pro-Replace is Pruna AI&#39;s character replacement model for putting the person from one or two photos into an existing video, keeping its camera, motion and soundtrack.</p>
<ul>
<li>Use <code>prunaai:p-video@2-pro-replace</code> with a source video in <code>inputs.video</code> and one or two photos in <code>inputs.referenceImages</code>.</li>
<li>Accepts source videos from 1 to 120 seconds and up to 8K, with aspect ratios from 1:4 to 4:1, and returns the same length and shape at 24 fps.</li>
<li>Generates at 480p or 768p, or upscales to 1080p, 2K or 4K.</li>
<li>Matches two photos to the people on screen from left to right, and takes an optional instruction of up to 4,000 characters to change that mapping or replace one person in a group.</li>
<li>Keeps the source soundtrack and syncs the new person&#39;s lips to it, with <code>settings.sourceAudioSync</code> and <code>settings.preserveAudio</code> to turn either part off.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Hosted OpenClaw Deprecation]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-04-hosted-openclaw-deprecated</guid>
        <category><![CDATA[deprecation]]></category>
        <pubDate>Wed, 04 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Hosted OpenClaw is deprecated and will be shut down on <strong>November 18, 2026</strong>. On that date, every hosted OpenClaw instance is deleted along with its workspace.</p>
<section id="migration"><h3 id="migration">Migration</h3><ul>
<li>Copy anything you want to keep from your instance&#39;s workspace before November 18.</li>
<li>Run OpenClaw on your own infrastructure and keep using Runware models through the <a href="/docs/tools/compatible-apis">compatible APIs</a>, by setting Runware&#39;s base URL and your API key in its model provider.</li>
<li>Install the <a href="/docs/tools/skills">Runware skills</a> in your OpenClaw to keep generating images, video and audio from the agent.</li>
</ul></section>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Kimi K3.1 Pro]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-04-kimi-k3_1-pro</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Wed, 04 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Kimi K3.1 Pro is now available through Runware&#39;s <a href="/docs/tools/compatible-apis#openai-chat-completions">OpenAI-compatible layer</a>.</p>
<ul>
<li>Use <code>moonshotai:kimi@k3.1-pro</code> as the model value in OpenAI-compatible requests.</li>
<li>Supports text and image inputs, structured outputs, and tool calling for coding, 3D modeling, financial research, and agent workflows.</li>
<li>Provides a 1,048,576-token context window with up to 1,048,576 output tokens.</li>
<li>Reasoning is always on, with effort adjustable from low to max.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Kling VIDEO 4.0 Flash]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-04-kling-video-4_0-flash</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Wed, 04 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Kling VIDEO 4.0 Flash is now available on Runware as the 720p model of the Kling 4.0 family, covering text-to-video, image-to-video, reference-driven generation and prompt-based video editing.</p>
<ul>
<li>Use <code>klingai:kling-video@4.0-flash</code> to generate 3 to 20 second videos at 720p in 16:9, 1:1, 9:16 and 21:9.</li>
<li>Starts from a first frame image, or from text alone.</li>
<li>Accepts up to 10 reference images, 5 reference videos and 7 elements, and edits a source video from a prompt.</li>
<li>Prompts point at each input by position, as <code>&lt;&lt;&lt;image_1&gt;&gt;&gt;</code> and <code>&lt;&lt;&lt;video_1&gt;&gt;&gt;</code>.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Krea 3]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-04-krea-3</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Wed, 04 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Krea 3 is Krea&#39;s image model for text-to-image generation and for editing selected regions of an existing image.</p>
<ul>
<li>Use <code>krea:krea@3</code> to generate from text at 30 sizes from 1K to 4K, across ten aspect ratios from 1:1 to 21:9.</li>
<li>Accepts up to eight reference images, with the output set by <code>resolution</code> (<code>1K</code>, <code>2K</code> or <code>4K</code>) or by one of the same 30 sizes.</li>
<li><code>settings.editRegions</code> takes up to 24 boxes. Boxes on the first reference image mark the areas to edit, and boxes on the other reference images mark the objects to bring in.</li>
<li>Each box can carry a <code>label</code>, referenced from the prompt as <code>@label</code>, and a <code>note</code> with its own instruction.</li>
<li><code>settings.thinkingLevel</code> sets how much the model reasons before generating, <code>fast</code> or <code>high</code>.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Model router allowed and excluded models]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-03-model-router-model-lists</guid>
        <category><![CDATA[feature]]></category>
        <pubDate>Tue, 03 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>The model router can now be held to a set of models you approve, with <code>routing.allowedModels</code> and <code>routing.excludedModels</code>.</p>
<ul>
<li><code>routing.allowedModels</code> takes at least two distinct names and nothing outside the list is considered, so a request none of them can serve is refused.</li>
<li><code>routing.excludedModels</code> takes at least one name and every other model stays a candidate.</li>
<li>Each name is an AIR identifier or a model slug, and the two lists cannot be sent together.</li>
<li>A name the catalog does not publish is refused, with the name quoted in the message, so a typo cannot quietly shrink the set it was meant to describe.</li>
<li>Names are matched case-insensitively, so <code>bfl:2@1</code> and <code>BFL:2@1</code> are the same model and a list holding both is refused as a repeat.</li>
</ul>
<p>See the <a href="/docs/models-api/model-router">model router docs</a>.</p>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Topaz Labs Wonder 4]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-03-topazlabs-wonder-4</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Tue, 03 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Topaz Labs Wonder 4 is a generative image enhancement and upscaling model that restores and enlarges an image in a single pass.</p>
<ul>
<li>Use <code>topazlabs:wonder@4</code> with the <code>upscale</code> task type to upscale an image by a factor of 1 to 6, where 1 enhances it at its original size.</li>
<li>Combines denoising, deblurring, compression artifact removal and face recovery in one pass, with no settings to tune.</li>
<li>Produces outputs up to 256 megapixels and 32,000 pixels per side, so the largest accepted input shrinks as the factor grows.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[HeyGen Video 1.0 30-second clips and auto duration]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-02-heygen-video-1-0-duration</guid>
        <category><![CDATA[feature]]></category>
        <pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>HeyGen Video 1.0 generates clips up to 30 seconds long and can set the length from the dialogue in the prompt.</p>
<ul>
<li>Set <code>duration</code> up to 30 seconds for text-to-video and for image-to-video from a first frame, up from 15.</li>
<li>Requests with reference images, videos or audio, or with <code>settings.promptEnhancement</code> set to <code>disabled</code> or <code>ultraquality</code>, keep the 15-second maximum.</li>
<li>Set <code>duration</code> to <code>auto</code> to size the clip from the quoted dialogue in the prompt, at least 8 seconds.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[MiniMax H3 Highspeed and H3 Lite]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-11-02-minimax-h3-highspeed-lite</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>MiniMax H3 Highspeed and MiniMax H3 Lite are now available as faster and lower-cost variants in the H3 video model family.</p>
<ul>
<li><strong>H3 Highspeed</strong> (<code>minimax:h3@highspeed</code>) balances generation quality and speed for everyday content creation.</li>
<li><strong>H3 Lite</strong> (<code>minimax:h3@lite</code>) is the lightweight variant for concept previews, creative drafts and batch variations.</li>
<li>Both generate 4–15 second videos at 480p or 768p across landscape, portrait, square and ultrawide aspect ratios, from text or from a first frame with an optional last frame.</li>
<li>Both accept up to nine reference images, three reference videos and three reference audio clips for reference-driven generation.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Ideogram 4.5 Precise Edit region edits]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-10-29-ideogram-4-5-precise-edit-region</guid>
        <category><![CDATA[feature]]></category>
        <pubDate>Thu, 29 Oct 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Ideogram 4.5 Precise Edit can limit an edit to one region of the source with <code>settings.editRegion</code>, so sources larger than 4 megapixels keep their full resolution.</p>
<ul>
<li>Set <code>settings.editRegion</code> to a box of <code>[x1, y1, x2, y2]</code> pixel coordinates in <code>inputs.seedImage</code>. The box must lie inside the image, with each side at least 256 pixels, an aspect ratio between 1:6 and 6:1 and at most 4,194,304 pixels.</li>
<li>Set it to <code>auto</code> to fit the region around <code>inputs.maskImage</code>, which is then required. A mask sent with a box must sit inside it.</li>
<li>Everything outside the region comes back unchanged and the output keeps the source dimensions. Without a region, sources over 4,194,304 pixels come back scaled down.</li>
<li><code>inputs.seedImage</code> and <code>inputs.maskImage</code> now accept files up to 50 MB, and <code>inputs.seedImage</code> takes up to 67,108,864 pixels, such as 8192×8192.</li>
</ul>]]></description>
      </item>
    
      <item>
        <title><![CDATA[Ideogram Vector 1.0]]></title>
        <link>https://runware.ai/docs/changelog</link>
        <guid isPermaLink="false">2026-10-29-ideogram-vector-1-0</guid>
        <category><![CDATA[model]]></category>
        <pubDate>Thu, 29 Oct 2026 00:00:00 GMT</pubDate>
        <description><![CDATA[<p>Ideogram Vector 1.0 is Ideogram&#39;s vector model for generating SVG files from a prompt and converting existing images into SVG.</p>
<ul>
<li>Use <code>ideogram:vector@1.0</code> with the <code>vectorize</code> task type, sending either <code>positivePrompt</code> to generate an SVG or <code>inputs.image</code> to convert one.</li>
<li>Generates SVGs in 15 sizes from 1:3 to 3:1, up to 2048 pixels on the long side, at low, medium or high quality.</li>
<li>Accepts up to 5 reference images to guide generation, and follows the first image&#39;s aspect ratio when no size is set.</li>
<li>Converts an image at its own dimensions, scaled down to 2048 pixels on the long side, with a simple, detailed or automatic level of detail.</li>
</ul>]]></description>
      </item>
    
      </channel>
    </rss>