This is a very straight-forward and simple workflow that I use often for NSFW generations with Wan 2.2. Interpolation for 2x frames is included, and a save-image node to grab the final frame generated, if you'd like to save that photo and continue your video!
I use two separate merged models, WAN 2.2 Enhanced NSFW | SVI | for the High model, and WAN 2.2 Remix V3.0 for the Low model. I have found good results with camera and prompt adherence along with movement with this combination. Feel free to check out any of the content I've posted for an idea of the settings/values I use.
Here are the Civitai links for the two models I use:
(High): https://civarchive.com/models/2053259/wan-22-enhanced-nsfw-or-svi-or-camera-prompt-adherence-lightning-edition-i2v-and-t2v-fp8-gguf?modelVersionId=2477552
(Low): https://civarchive.com/models/2003153/wan22-remix-t2vandi2v?modelVersionId=2771407
Description
Final iteration of my 2 Sampler I2V workflow.
FAQ
Comments (7)
Why does the end result appear so sped up?
This probably has to do with either your FPS or Rife Interpolation settings! Try using 20 fps, 81 or 101 frames (always add an extra frame) for 4 or 5 seconds, then ensure the Rife Interpolation node is ON and set to 2x multiplier, I use Rife 49.
Without the Rife 49 node active, your video will only be 4-5 seconds of very fast movement. The rife node set to 2x will add an extra frame between every frame, essentially turning 101 frames into 202 frames, it will slow the entire video down 2x and feel much more natural.
hay alguna lora que mantenga la consistencia de la cara inicial?
@juanra479 No, no que yo use. Pero a lo mejor puedes encontrar uno o crear uno tú mismo.
@MortimusTheGreat estoy nuevo en esto, investigare de como crear una, se que sera una lucha, pero hare el intento.
@juanra479 https://www.reddit.com/r/StableDiffusion/ Aquí hay mucha información útil
@MortimusTheGreat gracias!