Zuntan

Sigmas: `1, 0.94, 0.85, 0.73, 0.55, 0.28, 0` High: `3steps` Low: `3steps` Shift: `4.5` Enhance-A-Video weight: 1 Fresca: low 1, high 1.25, cutoff 17 SeedGacha SSR Video > Upscaler x1.5 & Encode Geforce RTX 3060 12GB: 840px 840px, 81f Sigma: `1.0, 0.97, 0.94, 0.90, 0.85, 0.795, 0.73, 0.65, 0.55, 0.42, 0.28, 0.14, 0.0` steps: `12` startsteps: `10-8`(`2-4`steps) Shift: `6.75`(`4.5` x1.5) Enhance-A-Video weight: `1` Disable Fresca Enable `addnoisetosamples` - wan2.2i2vhighnoise14Bfp16.safetensors - Wan2.2-Lightning/Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1/highnoisemodel.safetensors x1.0 - wan2.2i2vlownoise14Bfp16.safetensors - Wan2.2-Lightning/Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1/lownoisemodel.safetensors x1.0

Wan 2.1 I2V, 4 steps / generation. Both 480p and 720p LoRA are available. - `Wan21-BothI2v-FastMix2-14B`: Basic model. - `Wan21-BothI2v-FastPusaMix2-14B`: Pusa merged. - `Wan21-BothI2vAnisora-FastMix2-14B`: Anisora merged. Reduces the realistic appearance of the face. - `Wan21-BothI2vAnisora-FastPusaMix2-14B`: Pusa merged. Q3KM, Scheduler: lcm, Step: 4, Cfg: 1, Shift: 8, baseprecision: fp16fast, Use compile & blockswap HiresFix: Latent x1.5 bicubic, Denoise 0.45, Scheduler: lcm, Step: 3, Cfg: 1, Shift: 12, baseprecision: fp16fast, Use compile & blockswap - WanSelfForcingAcceleratorRank16.safetensors - lightx2vT2V14Bcfgstepdistillv2lorarank256bf16.safetensors - lightx2vI2V14B480pcfgstepdistillrank256bf16.safetensors Q3KM, Scheduler: flowmatchpusa, Step: 4, Cfg: 1, Shift: 8, baseprecision: fp16fast, Use compile & blockswap HiresFix: Latent x1.5 bicubic, Denoise 0.45, Scheduler: lcm, Step: 3, Cfg: 1, Shift: 12, baseprecision: fp16fast, Use compile & blockswap - WanSelfForcingAcceleratorRank16.safetensors - lightx2vT2V14Bcfgstepdistillv2lorarank256bf16.safetensors - lightx2vI2V14B480pcfgstepdistillrank256bf16.safetensors - Wan21PusaV1LoRA14Brank512bf16.safetensors Q3KM, Scheduler: lcm, Step: 4, Cfg: 1, Shift: 8, baseprecision: fp16fast, Use compile & blockswap HiresFix: Latent x1.5 bicubic, Denoise 0.45, Scheduler: lcm, Step: 3, Cfg: 1, Shift: 12, baseprecision: fp16fast, Use compile & blockswap - WanSelfForcingAcceleratorRank16.safetensors - lightx2vT2V14Bcfgstepdistillv2lorarank256bf16.safetensors - lightx2vI2V14B480pcfgstepdistillrank256bf16.safetensors - Q3KM, Scheduler: flowmatchpusa, Step: 4, Cfg: 1, Shift: 8, baseprecision: fp16fast, Use compile & blockswap HiresFix: Latent x1.5 bicubic, Denoise 0.45, Scheduler: lcm, Step: 3, Cfg: 1, Shift: 12, baseprecision: fp16fast, Use compile & blockswap - WanSelfForcingAcceleratorRank16.safetensors - lightx2vT2V14Bcfgstepdistillv2lorarank256bf16.safetensors - lightx2vI2V14B480pcfgstepdistillrank256bf16.safetensors - Wan21PusaV1LoRA14Brank512bf16.safetensors Scheduler: lcm, Step: 4, Cfg: 1, Shift: 8, baseprecision: fp16fast, Use compile & blockswap, EnhanceA(weight: 1, start: 0, end: 0) Hires. fix Latent x1.5 bicubic, Denoise 0.3, Scheduler: lcm, Step: 4, Cfg: 1, Shift: 12, baseprecision: fp16fast, Use compile & blockswap - Wan21T2V14Blightx2vcfgstepdistilllorarank32 - Wan21CausVid14BT2Vlorarank32v15nofirstblock - Wan2.1I2V14BFusionXLoRA - Wan21CausVid14BT2Vlorarank32v2 Scheduler: unipc, Step: 4, Cfg: 1, Shift: 8, baseprecision: fp16fast, Use compile & blockswap - Wan21-VACE-14B-fp16 - Wan21T2V14Blightx2vcfgstepdistilllorarank32 - Wan21CausVid14BT2Vlorarank32v2 - Wan21AccVidT2V14Blorarank32fp16 - Wan21CausVid14BT2Vlorarank32v15nofirstblock - wan2.1i2v480p14Bfp16.safetensors - Wan21-SkyReels-V2-I2V-14B-540Pfp16.safetensors - wan2.1i2v720p14Bfp16.safetensors - Wan21-SkyReels-V2-I2V-14B-720Pfp16.safetensors

EasyForge の fluxtool で、iNiverse Mix Flux v1.1 を Hyper Flux で 8steps 生成にして、GGUF と nf4 に変換したモデルです。 Geforce RTX 3060 12GB で、速度重視の nf4 なら 30秒、高品質の Q80 なら 45秒で生成できます。 Flux Dev 用 の LoRA も使えます。 1. `ExtractLora-FluxD-fp16.bat` を実行して、iNiverse から LoRA を Dim `128` で抽出 2. `ModelMergeLora.bat` を実行して、flux1-dev-hyper8.safetensors に抽出した LoRA を重み `1.0` でマージ 3. `ConvertGguf.bat` を実行して、マージモデルの `Q80`, `Q6K`, `Q4KS` に変換 4. Forge の `Checkpoint Merger` で `nf4` と `nf4-AIO(All-in-One, t5xxlfp16, ae, clipl 込み)` に変換 - `iNiverseV11x05-8step.` は 2. の LoRA 重みを `0.5` にしたものです。 - `iNiverseV11x05-FCAnimeV33x05-8step.` は、更に Flat Color Anime v3.3 を重み `0.5` でマージしたものです。 - プロンプトに「fcastyle anime」を追加してご利用ください。

- wan2.1i2v720p14Bfp16.safetensors - Wan21-SkyReels-V2-I2V-14B-720Pfp16.safetensors - Wan21CausVid14BT2Vlorarank32.safetensors T2v-SkyRel-CausVid05-14B-720p - Wan21-SkyReels-V2-I2V-14B-720Pfp16.safetensors - Wan21CausVid14BT2Vlorarank32.safetensors - wan2.1i2v480p14Bfp16.safetensors - Wan21-SkyReels-V2-I2V-14B-540Pfp16.safetensors - wan2.1i2v720p14Bfp16.safetensors - Wan21-SkyReels-V2-I2V-14B-720Pfp16.safetensors - Wan21T2V14Blightx2vcfgstepdistilllorarank32 - Wan2.1I2V14BFusionXLoRA - Wan21CausVid14BT2Vlorarank32v15nofirstblock - Wan21CausVid14BT2Vlorarank32v2 - Wan21T2V14Blightx2vcfgstepdistilllorarank32 - Wan21CausVid14BT2Vlorarank32v2 - Wan21AccVidT2V14Blorarank32fp16 - Wan21CausVid14BT2Vlorarank32v15nofirstblock

HyperSDXL(CreativeML Open RAIL++-M) と DMD2(CC BY-NC-SA 4.0) をマージした NoobAI 用の高速化用 LoRA です。 `NoobAiVPredv065s` 用に `Hyper SDXL 1step x0.52` と `DMD2 4step x1.0` を SVD マージした高速化 LoRA です。 - `Sampling method`: `RES Solver` - `Schedule type`: `Beta` - `Steps`: `4` - `CFG Scale`: `1.0` - `Hires. fix`: `x1.5` Hires. fix 使用前提です。 - `Hires. fix steps`: `4` - `Hires. Upscaler`: `Latent (nearest-exact)` Latent 系以外は遅くなります。 - `Denoising strength`: `0.7` - `LoRA Weight`: `1.0`

- WanNsfwI2v-14B-480pfp8e4m3fn.safetensors - Wan-AI/Wan2.1-I2V-14B-480P に下記の LoRA をマージしたモデルです。 - WanNsfwI2vEnd-14B-480pfp8e4m3fn.safetensors - alibaba-pai/Wan2.1-Fun-V1.1-14B-InP に下記の LoRA をマージしたモデルです。 - 重み `0.06` - `animepiston` - `PovMissionaryT2vv11` - `animecumv101` - 重み `0.04` - `deepthroatblowjobv10` - `BulgePovMissionaryv11` - `Theyarehavingsidesexv10` - `nsfwsksInAndOutv014` - `CumshotT2vv10` - 重み `0.02` - `Fingeringv11` - `Cowgirlv11` - `PovBlowjobv11` - `FemaleOrgasme60`