anime-Hyper-VAE我只是把UnstableinkdreamV9.0-anime 加上 HyperSD和VAE,其他沒變,給自己用的,其實可以不需要下載然後因為hyper-sd 所以建議CFG 1.5~2 取樣方法 LCM Schedule type AutoSteps:８-20再多也沒啥意義anime-Hyper-VAEI just add HyperSD and VAE to UnstableinkdreamV9.0-anime , no other changes, for my own use, in fact, you do not need to download!And because of hyper-sdSo we suggest CFG 1.5~2.Sampling method LCMSchedule type AutoSteps:8-20 more steps is meaningless.如果你喜歡我的工作可以幫我買杯咖啡，謝謝If you like what I do, you can buy me a cup of coffee. Thanks.私の仕事を気に入っていただけたら、コーヒーを一杯おごってください。 ありがとうございます。<a target="_blank" rel="ugc" href="https://ko-fi.com/vicentma">https://ko-fi.com/vicentma</a><h1 id="9.0-nq5nup4gb">9.0</h1>It's just a minor tweak to make it more adaptable to all kinds of LORA.Photo can be transformed into anime by LORA.anime is also possible.A few of the LORA I made can be put to good use.<a target="_blank" rel="ugc" href="https://civarchive.com/models/271354">https://civarchive.com/models/271354</a>Specialised Negative<a target="_blank" rel="ugc" href="https://civarchive.com/models/417945?modelVersionId=465621">https://civarchive.com/models/417945?modelVersionId=465621</a>只是小幅的微調,讓他更能夠良好的適應各種LORAphoto其實可以靠LORA轉變畫風成為動漫化anime一樣也可以幾個也是我做的LORA可以善用<a target="_blank" rel="ugc" href="https://civarchive.com/models/271354">https://civarchive.com/models/271354</a>專用的Negative<a target="_blank" rel="ugc" href="https://civarchive.com/models/417945?modelVersionId=465621">https://civarchive.com/models/417945?modelVersionId=465621</a><h1 id="8.0-omqgprga6">8.0</h1>it, but I've updated it with LECO and I'm going to put up the photo version first.Because it is impossible to combine animation and reality, after 7.5 it will be split into three genres - balance/anime/photorealBut recently things have changed and with some new LECO training methods some of the concepts can be shaved off.so you can use some lora to chage the style of base modelsuch as my lora<h3 id="8.0-balance-tl61r4jf5">8.0-balance</h3>2D,2.5D,3D All can do wellBut it is better to be in 2.5D digital art or something like that in the direction of fineness, which can give you more surprises.Better detailing, and can be combined with various LORA variations.Just Try it<h3 id="8.5-photoreal-version-3qokclt95">8.5-photoreal version</h3>This is a very realistic model for photorealism and hyperrealism<h3 id="v7.5-anime-kxxzjx3ge">V7.5-Anime</h3>This is a very well in manga,anime....etc<a target="_blank" rel="ugc" href="https://civarchive.com/models/131676/anime-minimalist">https://civarchive.com/models/131676/anime-minimalist</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/131537/racial-slider">https://civarchive.com/models/131537/racial-slider</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/131491/comic-linearts">https://civarchive.com/models/131491/comic-linearts</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/131491/comic-linearts">https://civarchive.com/models/131491/comic-linearts</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/127174/photo-paint-slider">https://civarchive.com/models/127174/photo-paint-slider</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/127689/celluloid-photo">https://civarchive.com/models/127689/celluloid-photo</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/132532/leconegative">https://civarchive.com/models/132532/leconegative</a><h1 id="7.5-kq46yv8u0">7.5</h1>the April version of unstableinkdreamBecause it is impossible to combine animation and reality, after 7.5 it will be split into three genres - balance/anime/photoreal<h3 id="7.5-balance-e6eoyeln2">7.5-balance</h3>photo anime , movie,manga, photorealism 2D,2.5D,3DAll can do wellJust Try it<h3 id="7.5-photoreal-version-56vdx5mpy">7.5-photoreal version</h3>This is a very realistic model for photorealism and hyperrealism ,and is also a bit naughty<h3 id="v7.5-anime-o9mp71cbc">V7.5-Anime</h3>This is a very well in manga,anime....etc<h3 id="v7.3-in24ikyxb">V7.3</h3>Using the MBW plug-in to remerge the model more detaile and style photoreal<h3 id="v7.0-py1rn0947">V7.0</h3>Not uploaded for a long time Using the auto-MBW plug-in to remerge the model Compared to V6 this is a more realistic model, but if the prompts are used properly it can also be used to produce anime. To be honest, the pure models have not improved much since mid-January, most of the models are already adequate, controlling the network and LORA is more important. The example is a quick draw of 45 cards using wildcrd and dynamicprompts, selected at random. Why is it a lot of fantasy WAIFU style? I like it ...... But the other styles actually look ok too I'll leave it to you to try it out<h3 id="v6-nrb4cg1to">V6</h3>Reconstruction the model to make the prompts itself can be used to create a variety of styles including realistic photos, realistic drawings, and animations like last three figmodel tag like : nvinkpunk,kuvshinov,dreamlikeart,samdoesart and modelshoot style analog stylewill effectiveand normal tag like : unreal engin 5,octane render, limited palette flat color,Hyperrealism,Photorealism photorealistic also worked wellsome negative prompt was come from civitai and huggingface google can find itV5photoreal- updatedTo be honest, I'm sorry, this should come out together with yesterday's V5, one is more towards animation and the other more towards reality But has not been adjusted very satisfactory, the effect is actually a bit close to V4If you want to be more photo-realistic, please add (photorealistic) at the begin , and real person's name will effectIf you want to favor animation, please add (nvinkpunk,kuvshinov,dreamlikeart,samdoesart,modelshoot style,) at the end.........not only anime style but girls jump out lolV5 updatedEnhances cartoonish and stylized intensity with better mastery of the facemodel tag like : nvinkpunk,kuvshinov,dreamlikeart,samdoesart and modelshoot style will effectiveand normal tag like : unreal engin 5,octane render, limited palette flat color,Hyperrealism,Photorealism also worked wellneed nsfw in nagtive promptsV4 updatedDont ask me where are v2 or v3,mix a lot of model ,v4 is the best onemix by V1,RPG,Protogen,dreamlikesamkuvshinov,inkpunk.......model tag like : nvinkpunk,kuvshinov,dreamlikeart,samdoesart and modelshoot style will effectiveand normal tag like : unreal engin 5,octane render, limited palette flat color,Hyperrealism,Photorealism also worked wellUnstableSamInkDream DiffusionCombined from the following modelsA:Unstable PhotoReal v0.5<a target="_blank" rel="ugc" href="https://www.reddit.com/r/StableDiffusion/comments/zhg18s/unstable_diffusion_here_were_excited_to_announce/">https://www.reddit.com/r/StableDiffusion/comments/zhg18s/unstable_diffusion_here_were_excited_to_announce/</a>B:inkpunk<a target="_blank" rel="ugc" href="https://huggingface.co/Envvi/Inkpunk-Diffusion">https://huggingface.co/Envvi/Inkpunk-Diffusion</a>C:dreamlikesamkuvshinov<a target="_blank" rel="ugc" href="https://civarchive.com/models/1473/dreamlikesamkuvshinov">https://civarchive.com/models/1473/dreamlikesamkuvshinov</a>A0.5+B0.5=D D0.6+C0.4＝UnstableSamInkDreamTrigger Words:nvinkpunk,dreamlikeart,samdoesart,kuvshinov,

UnstableInkDream

To be honest, I'm sorry, this should come out together with yesterday's V5, one is more towards animation and the other more towards reality But has not been adjusted very satisfactory, the effect is actually a bit close to V4 If you want to be more photo-realistic, please add (photorealistic) at the beginIf you want to favor animation, please add (nvinkpunk,kuvshinov,dreamlikeart,samdoesart,modelshoot style,) at the end.........not only anime style but girls jump out lolAs chinese new year will come , so maybe not update models next two weeks to have a vacation. In addition to the example of the problem 1. example vae use vae-ft-mse-840000-ema-pruned.vae 2. The previous examples I will follow my habit of using very long tokens, but it seems that the site may not be able to let too long tokens all show, so this time especially changed the short point, the effect is also okay........maybe 3. I will also put the original picture of the example on my HUGGINGFACE and the test prompts file used this time <a target="_blank" rel="ugc" href="https://huggingface.co/Oosayam/UnstableSamInkDream/tree/main">https://huggingface.co/Oosayam/UnstableSamInkDream/tree/main</a>4. In addition, please note that even if all the prompts , models ,hypernetworks ,VAEs and seed are all the same, with the version of the WEBUI ,GPU, and the use XFORMERS or not will all have an impact........this may not so bad though.........

Files with SHA256: 199638b9b2430e386f9fb0de096dc6524f11a9fb050bf3903decd47732d8f2da