Loading
Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
This is old stuff. AnimateDiff was one of the first Video AI solutions for offline needs. And this is the most basic workflow for it. The base for other AnimateDiff workflows.
Runtime profile
Source description
Howto
Adjust prompt. Generate.
Description
This is old stuff. AnimateDiff was one of the first Video AI solutions for offline needs. And this is the most basic workflow for it. The base for other AnimateDiff workflows.
This workflow here is about prompt traveling with AnimateDiff. You choose different prompts for different keyframes. And in the video the prompts morphs into each other.
There are two workflows available. One for AnimateDiff 2 and 3. And one for the AnimateLCM, which is based at AnimateDiff 2.
AnimateDiff and Stable Diffusion 1.5, which are used in conjunction, are trained at 512 pixels in size. So you better won't go this much higher in creation size. The result will become inconsistent. What you can try is to upscale afterwards.
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
Use this workflow to make an AnimateDiff video by assigning different prompts to keyframes; the video transitions between those prompts.
Choose either the AnimateDiff 2/3 version or the AnimateLCM version, which is based on AnimateDiff 2.
Both versions include upscaling; the creator's example is a six-second, 1280 × 720 video made with simple AI upscaling.
Version v1.0 is listed for the SD 1.5 base model and published by Arunderan.
For the AnimateDiff 2/3 version, the listed model requirements are analogMadness_v70.safetensors, RealESRGAN_x2.pth, rife47.pth, v2_lora_ZoomIn.ckpt, v3_sd15_adapter.ckpt, v3_sd15_mm.ckpt, and vae-ft-mse-840000-ema-pruned.safetensors.
For the AnimateLCM version, the listed model requirements are AnimateLCM_sd15_t2v.ckpt, AnimateLCM_sd15_t2v_lora.safetensors, RealESRGAN_x2.pth, realismBYSTABLEYOGI_v6LCM .safetensors, rife47.pth, v2_lora_ZoomIn.ckpt, and vae-ft-mse-840000-ema-pruned.safetensors.
The creator notes AnimateDiff and SD 1.5 were trained at 512 pixels and advises against generating much larger, as results may become inconsistent.
Suggestion · not verified
For upscaling, the creator recommends using a separate workflow with a second KSampler.
Suggestion · not verified
The creator says the workflows were made with 8 GB of VRAM, notes repeated out-of-memory issues, and recommends at least 12 GB for proper upscaling.
Suggestion · not verified
No license information is listed for this workflow.
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
3 excerptsSource context: 412 downloads · Type Workflows · Base model SD 1.5
Both workflows comes with upscaling. But i recommend to upscale in an extra workflow. By a second Ksampler. See this article here:
Time
The example with a length of six seconds and a resolution of 1280 x 720 (simple upscaling with AI) finished in two and a half minutes at an 4060 TI with 16gb Vram. It does not use the best upscaling method though.
Requirements
These workflows were generated with 8 gb vram back in the days. However, i ran too often into an OOM. And when you want to do a proper upscaling then i highly suggest a card with at least 12 gb vram.
Requirements
LoRA · 128 MB · SAFETENSORS · Unknown
Checkpoint · 1.69 GB · CKPT · fp32 · Unknown
Upscaler · 64.0 MB · PTH · Unknown
realismBYSTABLEYOGI_v6LCM .safetensors
Not resolvedCheckpoint · Unknown
rife47.pth
File unverifiedCheckpoint · Unknown
LoRA · 73.9 MB · PICKLETENSOR · Unknown
vae-ft-mse-840000-ema-pruned.safetensors
Not resolvedVAE · Unknown
Source context: 369 downloads · Type Workflows · Base model SD 1.5
Source context: 353 downloads · Type Workflows · Base model SD 1.5
