Wan 2.2 S2V-14B in ComfyUI Workflow: Speech-to-Video (GGUF, Low VRAM) скачать в хорошем качестве

Wan 2.2 S2V-14B in ComfyUI Workflow: Speech-to-Video (GGUF, Low VRAM)

Wan 2.2 Speech-to-Video just landed, and I’m running the full InfinityTalk-style ComfyUI workflow end-to-end so you can see Wan 2.2 Speech-to-Video in action on real clips. In this video, you'll learn: • How I set up the Wan 2.2 S2V-14B model in ComfyUI (clean folder paths, exact nodes) • A simple image-to-video workflow that keeps lip sync tight, with steps/CFG and sampler settings • What changes when I swap to GGUF for low-VRAM cards, and how speed vs quality compares I keep it practical. One still photo and an audio track become a talking clip. First demo: singing by the ocean. Second demo: a sofa dialogue shot. You’ll see the mouth hit vowels and consonants on time, natural blinks, and small head motion that feels right. I also show the quick switch between Lightning (4 steps, CFG 1) and a slower pass without the LoRA, plus a low-VRAM run on Wan 2.2 GGUF. If you’ve used Wan 2.1 in ComfyUI before, this will feel familiar; the same text encoder and VAE work here too. I repeat the key settings on screen so you can copy them without pausing a dozen times. If you’re new to this, don’t worry—everything is straight to the point. File Need: •wan2.2_s2v_14B_bf16.safetensors •wav2vec2_large_english_fp16.safetensors • Free Workflow: https://aistudynow.com/wan-2-2-s2v-14... #Wan22 #ComfyUI #SpeechToVideo