Wan 2.2 5B versus 14B: match the task, model stages and VAE

File and download boundary: ComfyUI World provides source references and metadata only. Model and LoRA weights, node packages, and workflow files stay with their original providers; this site does not store, mirror, or proxy them. Downloads and access conditions are controlled by each provider.

Purpose and scope

Distinguish 5B TI2V from the 14B text/image-to-video routes. A shared Wan 2.2 label does not make weights, VAE, encoder and latents freely interchangeable.

Source-supported context

Official instructions separate 5B and 14B, using wan2.2_vae for the 5B example and wan_2.1_vae for listed 14B routes. Some 14B workflows use high/low-noise model stages. This record's table contains only three documented 5B files. Source 1 Source 2

Before you start

1. 5B TI2V has text/image input paths belonging to its corresponding template. Check that your ComfyUI build includes the example's core nodes; the official guide notes that some may not yet be in the latest stable release.

2. 14B T2V and I2V have separate task, model and stage requirements.

3. Acceleration, quantization and third-party wrappers are not universal replacements for arbitrary native components.

Steps and checks

The sequence below is an editorial procedure based on the cited context, not a diagnosis already confirmed for your environment.

Step 1. Choose the task-specific 5B, 14B T2V or 14B I2V template rather than downloading based on filename appearance.

Step 2. Record diffusion, encoder, VAE and latent construction with each file source/directory.

Step 3. For dual-stage graphs, preserve the high/low-noise weights and sampling boundary instead of selecting one file in both loaders.

Step 4. Begin with the selected template's supported dimensions and frame count. If the workflow and model support a smaller load, change one setting at a time and inspect sampling and decoding pressure separately.

Completion check

Loaders and input mode match the selected route and the final video saves. A complete file list is not proof of minimum VRAM.

Limitations and cautions

Upstream hardware guidance is not this site's benchmark. No 5B/14B weights or quantization quality/speed comparisons were tested here.

Files named by the source documentation

These are the selected documentation example's declarations, not inspected local files or direct downloads. No weights or checksum results are included; verify the revision for your task.

File	Directory category	Documented source
wan2.2_ti2v_5B_fp16.safetensors	diffusion_models	Source
umt5_xxl_fp8_e4m3fn_scaled.safetensors	text_encoders	Source
wan2.2_vae.safetensors	vae	Source

Original sources

• Wan2.2 native workflows — original record checked 2026-09-21.
• ComfyUI model troubleshooting — original record checked 2026-09-21.

Localization prepared: 2026-09-22. The source-check dates above were inherited from the existing article; this translation does not claim they were all rechecked today. No GPU run, minimum-memory measurement or runtime guarantee is supplied.

Related troubleshooting and guides

• First Wan 2.2 5B image-to-video task: verify the input through to the saved clip
• VAE and latent channel mismatch: verify the exact variant, not just its family name
• VAE out of memory after sampling finishes: diagnose decoding separately