EasyAnimate Sampler for Image to Video

EasyAnimateI2VSampler is image-to-video for the older EasyAnimate weights: V3 and V4. If you're on the current V5.1 line you want the V5 sampler of the same name instead, and the pack makes the distinction easy - the V5 nodes are literally named EasyAnimateV5_I2VSampler. Don't treat these as interchangeable. The legacy samplers handle 144-frame clips at up to 24fps (V4's signature), while V5.1 is capped at 49 frames at 8fps but is a much better model.

This node is a single bundled sampler: you give it the loaded model, a prompt, and an image, and it returns the finished video frames as an IMAGE batch. That output is what you feed VideoHelperSuite to write an actual .mp4 - which is why the pack's install instructions make you clone it.

The inputs that matter

• easyanimate_model - from LoadEasyAnimateModel, loaded with the matching V3/V4 -InP weights (and the matching config). You do not use a -Control checkpoint here.
• prompt and negative_prompt - both STRING_PROMPT type, meaning they're wired from a TextBox/EasyAnimate_TextBox node rather than typed inline. EasyAnimate is bilingual; English works fine.
• start_img and end_img (optional) - the image inputs. start_img is the first frame; add end_img and the model will try to end on that target frame. This is a genuinely useful EasyAnimate feature and most workflows only use the start.
• video_length - 8 to 144 in steps of 8, default 72. For V3/V4's 24fps models, 144 frames is a 6-second clip.
• base_resolution - 512/768/960/1024, default 768. This is the clever bit: the node reads your start image's aspect ratio, scales it to the nearest ratio at your chosen base resolution, and rounds to multiples of 16. You don't set width/height here - you set a quality target and it handles the framing.
• seed, steps, cfg - the usual suspects. 25 steps and CFG 7 are the sane defaults.
• scheduler - Euler, Euler A, DPM++, PNDM, or DDIM (default). No Flow option here - that's a V5.1 thing.

The output

images - an IMAGE tensor, (frames, height, width, channels). Wire it into a VideoHelperSuite video combine node to get a file. If you're feeling bold you can also treat the frames as a batch and do post-processing.

Reality check

The community was lukewarm on EasyAnimate's motion even at V5.1's launch - "basic motion, mostly slowmo-like" was a common verdict - and V3/V4 are older and worse. These legacy samplers are really for people who already have V3/V4 checkpoints (they're lighter than the 12B, which has appeal on modest GPUs) or who want the 144-frame capability. For everyone else, load EasyAnimateV5.1-7b-zh-InP and use EasyAnimateV5_I2VSampler. The install is the same either way: clone the pack or grab it via ComfyUI Manager, put weights in ComfyUI/models/EasyAnimate/, and make sure VideoHelperSuite is around for saving.