{"id":18197,"date":"2026-11-17T16:09:04","date_gmt":"2026-11-17T07:09:04","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/minimax-h3-image-character-swap-2\/"},"modified":"2026-11-17T16:09:04","modified_gmt":"2026-11-17T07:09:04","slug":"minimax-h3-image-character-swap-2","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/minimax-h3-image-character-swap-2\/","title":{"rendered":"MiniMax-H3-Image-Character-Swap Image Editing Model: 4GB+ VRAM"},"content":{"rendered":"<p><em>Sample outputs are available on the <a href=\"https:\/\/huggingface.co\/akhaliq\/MiniMax-H3-Image-Character-Swap\">model card<\/a>.<\/em><\/p>\n<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/akhaliq\/MiniMax-H3-Image-Character-Swap\">akhaliq\/MiniMax-H3-Image-Character-Swap<\/a><\/td>\n<\/tr>\n<tr>\n<td>Family guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-minimaxai-minimax-h3-en\/\">MiniMax-H3 guide (14 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-minimax-en\/\">MiniMax: models and licenses<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-11-14<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>minimax-h3-image-community-license-agreement<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">safetensors<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Unverified (not confirmed by a primary source)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p><strong>Note: This news has not been officially confirmed and is based on unverified information.<\/strong> The following content is based on the details provided in the Hugging Face model card.<\/p>\n<p><code>akhaliq\/MiniMax-H3-Image-Character-Swap<\/code> is described as a checkpoint where <code>akatz-ai\/MiniMax-H3-Character-Swap-LoRA<\/code> has been merged directly into the <code>MiniMaxAI\/MiniMax-H3-Image<\/code> image DiT. By inputting a scene photo (<code>), a character reference image (<\/code>), and text specifying the person to be replaced, the model outputs an image with the character swapped while preserving the background, lighting, other people, and poses (image-to-image). According to the card, it also supports swapping anime-style characters into live-action scenes, which the base model reportedly fails to do.<\/p>\n<h2>Specifications<\/h2>\n<ul>\n<li>Base: <code>MiniMaxAI\/MiniMax-H3-Image<\/code>. The card states this is the first merged version that ports a character-swap LoRA trained for video into an image model.<\/li>\n<li>Merge Details: It is stated that <code>h3_character_swap_v1.5.safetensors<\/code> (rank 16, 312 modules, attention+FFN of 50 DiT blocks and 2 token-refiner blocks) was fused at a strength of 1.0.<\/li>\n<li>Base Model DiT (from the original model card): A 33.1B parameter single-stream transformer. It uses the MiniMax-H3 text encoder as-is.<\/li>\n<li>Output Resolution: The base model generates at approximately 16MP (4096&#215;4096 to 6272&#215;2688). This merged version&#8217;s card recommends a 4MP starting canvas. The examples use <code>width=2720, height=1536<\/code> (multiples of 32), and it is reported that 16MP also works.<\/li>\n<li>Recommended Settings: euler \/ simple, 6 steps, CFG 1, denoise 1. As it is a distilled model, no adjustments are required.<\/li>\n<li>Input: Two images (scene, then character). It is recommended to specify the person and include instructions in the prompt to &#8220;maintain face, hair, and clothing, while preserving background, lighting, camera, and other people.&#8221;<\/li>\n<li>Format: Only the <code>transformer_diffusers\/<\/code> layout of the modular pipeline is included. The traditional <code>MiniMaxH3ImagePipeline<\/code> layout is not included.<\/li>\n<li>License Terms (Summary from the card):<\/li>\n<li>Free for use, including commercial use, in all regions except the EU, UK, South Korea, and the US. &#8211; Written permission from MiniMax is required if annual revenue exceeds $1M or if provided as a hosted service to third parties. &#8211; No rights are granted under the community license in excluded regions. &#8211; MiniMax reportedly does not claim rights to the generated output. &#8211; The card itself states this is &#8220;not legal advice,&#8221; so checking the LICENSE file is necessary.<\/li>\n<\/ul>\n<h2>Performance<\/h2>\n<p>The following is the publisher&#8217;s (akhaliq) own evaluation and has not been verified by a third party.<\/p>\n<ul>\n<li>In a 3-way comparison (native, standalone LoRA, merged version) of 6 scenes x 2 characters x 3 seeds (4MP each), 17 scored comparisons were conducted. The merged version was chosen in 14\/17 (82%) of cases, the standalone LoRA in 2, and the native in 1.<\/li>\n<li>There was only one evaluator, and the card itself states this is &#8220;a signal of strong preference, not a benchmark.&#8221;<\/li>\n<li>The comparison dataset is reportedly available at <code>akhaliq\/h3-image-swap-tests<\/code>.<\/li>\n<li>Verification was performed on 8 difficult cases (profile, small face, occlusion, full body, group).<\/li>\n<\/ul>\n<p>The comparison table in the card is primarily image-based and contains no numerical data. The summary is as follows:<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Case<\/th>\n<th>Native H3-Image<\/th>\n<th>This Model<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Anime char to profile photo<\/td>\n<td>Original face remained (failed in 3\/3 seeds)<\/td>\n<td>Succeeded with same seed<\/td>\n<\/tr>\n<tr>\n<td>Live-action char to cafe<\/td>\n<td>Skin tone became brighter than reference<\/td>\n<td>Maintained identity<\/td>\n<\/tr>\n<tr>\n<td>Anime char to full-body jogging<\/td>\n<td>Pose changed to standing<\/td>\n<td>Maintained running pose and clothing<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>It is reported that the merged version suppresses the skin tone shifts and pose changes seen in the native model. On the other hand, the card lists the following limitations:<\/p>\n<ul>\n<li>The LoRA was trained for video, and its porting to images is limited to verification in 8 cases.<\/li>\n<li>In extreme profiles or combinations of different styles, the redrawn gaze direction may become unnatural.<\/li>\n<li>Deviations may occur in extreme poses, strong occlusions, or very small faces.<\/li>\n<li>Results depend on the quality of the reference image.<\/li>\n<\/ul>\n<p>It is specified that the orientation of the replaced person is adjusted to match the source person.<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<p>According to the model card, the goal is &#8220;image editing to replace an existing person with the face, hair, and clothing of a reference character.&#8221; These are all based on the publisher&#8217;s own description and are unverified.<\/p>\n<ul>\n<li>Cross-style swapping, such as replacing anime\/illustration-style characters into live-action scenes.<\/li>\n<li>Swapping into non-frontal faces, such as profiles. It targets cases where the base MiniMax-H3-Image failed in 3\/3 seeds.<\/li>\n<li>Editing where you want to keep the background, lighting, other people, and poses, such as conversations in a cafe or full-body jogging.<\/li>\n<li>Editing where you specify a person to replace, such as in a group photo.<\/li>\n<\/ul>\n<p>The original LoRA (<code>akatz-ai\/MiniMax-H3-Character-Swap-LoRA<\/code>) was trained for video-based swapping. This model is a derivative version that ports it to an image model and fuses it into a single checkpoint, aiming for use without additional LoRA files or nodes. The tags are <code>image-to-image<\/code> and <code>image-editing<\/code>.<\/p>\n<p>It is stated that a clean multi-view character sheet is best for the reference image. When replacing real people, the card requests consideration for consent and portrait rights.<\/p>\n<h2>How It Differs from Similar Models<\/h2>\n<p>As an article in the same MiniMax-H3 series, there is <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/pika-h3-flex\/\">&#8220;Pika-H3-Flex&#8221; Video Generation Model with Audio: List of Distributed Files<\/a>. This is a LoRA adapter aimed at generating videos with synchronized stereo audio. This model differs in that it specializes in character replacement for still images and has the LoRA already fused into the checkpoint.<\/p>\n<p><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/fasth3-omni-comfy\/\">&#8220;FastH3-Omni-Comfy&#8221; Video\/Audio Generation Model with Image\/Audio Reference: Required VRAM 80GB+ \/ List of Distributed Files<\/a> is a Reference-to-Video system that generates video and audio simultaneously using image, video, and audio as references. This model is for editing, taking two images as input and outputting one image, so the use case is different.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>4GB (laptop iGPU \/ phone class)<\/td>\n<td>Original precision<\/td>\n<td>0.0GB<\/td>\n<td>0.0GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model&#8217;s authors. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p>The publisher distributes this model as safetensors.<\/p>\n<p><strong>License \u2014 <code>minimax-h3-image-community-license-agreement<\/code>:<\/strong> A custom license from the publisher. Check the original terms directly, including whether commercial use is permitted.<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats and the license field. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:lab --><\/p>\n<h2>Our Own Measurements<\/h2>\n<h3>Measurements We Did Not Take<\/h3>\n<p>We have not confirmed that this site meets the commercial-use terms of this model&#8217;s license (<code>minimax-h3-image-community-license-agreement<\/code>). Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<p><!-- \/lmw:lab --><\/p>\n<p><!-- lmw:files --><\/p>\n<h2>Distributed Files<\/h2>\n<p><em>Weight files published in <a href=\"https:\/\/huggingface.co\/akhaliq\/MiniMax-H3-Image-Character-Swap\/tree\/main\">akhaliq\/MiniMax-H3-Image-Character-Swap<\/a>, listed by this site from the Hugging Face API. Sizes are the actual file sizes.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>File<\/th>\n<th>Size<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><code>comfyui\/minimax_h3_image_character_swap_bf16.safetensors<\/code><\/td>\n<td>40.23GB<\/td>\n<\/tr>\n<tr>\n<td><code>lora\/h3_character_swap_v1.5.safetensors<\/code><\/td>\n<td>155MB<\/td>\n<\/tr>\n<tr>\n<td><code>lora\/h3_image_character_swap_diffusers.safetensors<\/code><\/td>\n<td>173MB<\/td>\n<\/tr>\n<tr>\n<td><code>transformer_diffusers\/diffusion_pytorch_model-*-of-00007.safetensors<\/code><\/td>\n<td>66.29GB (7 split files)<\/td>\n<\/tr>\n<tr>\n<td><code>vae_diffusers\/diffusion_pytorch_model.safetensors<\/code><\/td>\n<td>2.34GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:files --><\/p>\n<h2>How to Get It<\/h2>\n<ul>\n<li>The distribution format is safetensors. A modular pipeline layout for diffusers (<code>transformer_diffusers\/<\/code>) and a single file for ComfyUI are provided. There is no gated setting requiring license agreement.<\/li>\n<li>The repository is <code>akhaliq\/MiniMax-H3-Image-Character-Swap<\/code>.<\/li>\n<li>For diffusers, <code>diffusers==0.43.0<\/code> is required. The text encoder references the <code>MiniMaxAI\/MiniMax-H3-Image<\/code> side and is loaded automatically.<\/li>\n<\/ul>\n<pre><code class=\"language-python\">from diffusers import ModularPipeline\nREPO = &quot;akhaliq\/MiniMax-H3-Image-Character-Swap&quot;\nblocks = ModularPipeline.from_pretrained(REPO).blocks\npipe = blocks.init_pipeline(REPO)\npipe.load_components(dtype=torch.bfloat16)\n<\/code><\/pre>\n<ul>\n<li>Images are passed in the order <code>image=[scene_image, character_image]<\/code> (scene, then character). <code>width<\/code> and <code>height<\/code> should be multiples of 32.<\/li>\n<li>According to the card, running the whole system on a single GPU requires approximately 140GB of VRAM (DiT 66GB + Qwen3-VL encoder 67GB). If the encoder is run remotely, only the DiT and VAE need to be on the GPU.<\/li>\n<li>ComfyUI requires 0.45.0 or higher and template pack 0.11.104 or higher. A dedicated workflow JSON is in the repository, and it can also be used with the official &#8220;MiniMax-H3-Image: Image Edit&#8221; template. Point the diffusion model loader to <code>minimax_h3_image_character_swap_bf16.safetensors<\/code>. Use the text encoder and VAE from <code>Comfy-Org\/MiniMax-H3-Image<\/code>.<\/li>\n<li>According to the card, verification of the ComfyUI version is limited to the same calculations confirmed via diffusers but in a different layout. If there are any ComfyUI-specific issues, you are asked to report them in the discussions.<\/li>\n<li>If you want to adjust the strength, the standalone LoRA (under <code>lora\/<\/code>) is also distributed in the same repository.<\/li>\n<\/ul>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omni-comfy-released\/\">FastH3-Omni-Comfy: 80GB+ VRAM, File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fastvideo-fasth3-trim-comfy-released\/\">FastVideo-FastH3-Trim-Comfy Text-to-Video and Audio Model: File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omni-fp8-comfy-released-2\/\">FastH3-Omni-FP8-Comfy Audio-Synced Video Generation Model: 32GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omni-nvfp4-qad-comfy-released\/\">FastH3-Omni-NVFP4-QAD-Comfy: 16GB+ VRAM, File List<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Find models by VRAM<\/strong> (This model runs from the 4GB tier) \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-8gb-en\/\">Other models that run on a 8GB GPU<\/a><\/li>\n<li><strong>Explore the same model family<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-minimaxai-minimax-h3-en\/\">MiniMax-H3 family overview (14 articles, 4 converted builds)<\/a><\/li>\n<li><strong>Formats this model is available in<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">Safetensors format guide and models<\/a><\/li>\n<li><strong>Learn about the publisher<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-minimax-en\/\">MiniMax: models, licenses and articles<\/a><\/li>\n<li><strong>Other models for the same task<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/models-by-task-en\/#task-image\">Other image generation models<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/akhaliq\/MiniMax-H3-Image-Character-Swap\">https:\/\/huggingface.co\/akhaliq\/MiniMax-H3-Image-Character-Swap<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/MiniMaxAI\/MiniMax-H3-Image\">https:\/\/huggingface.co\/MiniMaxAI\/MiniMax-H3-Image<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/pika-h3-flex\/\">https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/pika-h3-flex\/<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/fasth3-omni-comfy\/\">https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/11\/17\/fasth3-omni-comfy\/<\/a><\/li>\n<\/ul>\n<blockquote>\n<p><strong>This article contains unverified information.<\/strong> We will append an update note once it is confirmed by a primary source.<\/p>\n<\/blockquote>\n","protected":false},"excerpt":{"rendered":"<p>A look at the MiniMax-H3-Image-Character-Swap checkpoint, which merges a character-swap LoRA into the MiniMax-H3-Image DiT for image-to-image editing.<\/p>\n","protected":false},"author":1,"featured_media":18196,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[203],"tags":[5234,5236,677,2742,730,4846,209,1565,1839],"class_list":["post-18197","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-image-video-and-audio","tag-akhaliq-en","tag-akhaliq-minimax-h3-image-character-swap-en","tag-comfyui-en","tag-image-to-image-en","tag-minimax-h3-en","tag-minimax-h3-image-en","tag-safetensors-en","tag-unverified","tag--en"],"lang":"en","translations":{"en":18197,"ja":18195},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/18197","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=18197"}],"version-history":[{"count":0,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/18197\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/18196"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=18197"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=18197"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=18197"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}