Wan2.2 14B Rapid AllInOne and Wan2.1 T2V 1.3B Diffusers: what they are and what they need

Wan2.2 14B Rapid AllInOne is a mix of Wan 2.2 made by Hugging Face user Phr00t, and Wan2.1 T2V 1.3B Diffusers is Wan-AI’s official text-to-video checkpoint (Apache 2.0, 8.19 GB VRAM), while this site currently delivers images only, not video: its photo editor edits one photo in about 37 seconds for 10 credits. Free with sign-up: 10 credits = 1 image, then from $0.13 per image.

Your edited photo appears here (this site’s editor)
Studio background edit made on this site: the same woman on a plain light-grey studio backdrop
Input photo (AI-generated, fictional): a smiling woman with curly brown hair in a green sweater, in a parkYour photo
Type your own change, or leave this empty and pick a one-tap edit below.0/400

Style changes (cartoon, 3D, anime) are not reliable yet. If the result looks unchanged, try again or describe the change more specifically.

One-tap edits

  • Images only: this site does not make or cut video
  • Not affiliated with Wan-AI or Phr00t
  • Free with sign-up: 10 credits = 1 image
VRAM stated for Wan2.1 T2V 1.3B (model card)
8.19 GB
VRAM stated for Wan 2.2 14B on one GPU (README)
80 GB
license on all three Hugging Face pages
Apache 2.0
what this site delivers; no video
Images only

What Wan2.1 T2V 1.3B Diffusers and Wan2.2 14B Rapid AllInOne are

Wan2.1 T2V 1.3B Diffusers (official, Wan-AI)

  • Wan-AI/Wan2.1-T2V-1.3B-Diffusers is a text-to-video checkpoint from the Wan team, in the Diffusers format. Its card says the weights and inference code were released on Feb 25, 2025, and the license is Apache 2.0. The card also says the license “must not involve sharing any content that violates applicable laws.”Source: Hugging Face model card ·
  • The card says the T2V-1.3B model “requires only 8.19 GB VRAM”. 480P is the supported size (832 x 480 in its commands). It says the model can also make 720P, but less stably, because it was trained at lower resolution.Source: Hugging Face model card ·
  • Its Diffusers example installs diffusers, transformers and accelerate, loads WanPipeline, then asks for height 480, width 832, 81 frames, guidance scale 5.0 and exports at 15 fps, which the card describes as a 5-second video. A second route uses the repo’s generate.py with --task t2v-1.3B, and offload flags to save memory.Source: Hugging Face model card ·
  • The Diffusers docs list this checkpoint among the supported Wan checkpoints, with a WanPipeline example and memory options such as model CPU offload.Source: Diffusers docs ·
  • The Wan 2.1 README news says Diffusers support for text-to-video and image-to-video was added on Mar 3, 2025 and ComfyUI support on Feb 27, 2025. The README’s own to-do list still shows both as unchecked, so two parts of that page disagree.Source: Wan 2.1 README ·

Wan 2.2 14B (official, Wan-AI)

  • Wan 2.2’s code and weights were released on Jul 28, 2025, under Apache 2.0. The text-to-video and image-to-video A14B models use a two-expert mixture of experts: a high-noise expert for early steps and a low-noise expert for detail. Each expert is about 14B parameters, about 27B in total, with 14B active at each step. They support 480P and 720P.Source: Wan 2.2 README ·
  • The README says the single-GPU command for the A14B models “can run on a GPU with at least 80GB VRAM”, and lists --offload_model True, --convert_model_dtype and --t5_cpu for out-of-memory cases. It says the smaller dense TI2V-5B needs at least 24GB (for example an RTX 4090).Source: Wan 2.2 README ·
  • The Diffusers card for the T2V A14B model shows 720P (1280 x 720), 81 frames at 16 fps, 40 steps and WanPipeline. It says the example needs the Diffusers main branch, not the PyPI release.Source: Hugging Face model card ·

Wan2.2 14B Rapid AllInOne (Phr00t on Hugging Face)

  • The repo Phr00t/WAN2.2-14B-Rapid-AllInOne sits under a personal Hugging Face account, not under Wan-AI. The card describes “mixtures of WAN 2.2 and other WAN-like models and accelerators”, with CLIP and the VAE included, in FP8 precision; its model tree lists Wan-AI/Wan2.2-I2V-A14B as the base. Its license tag is apache-2.0.Source: Hugging Face model card ·
  • How to use it, per the card: put the single file in ComfyUI’s checkpoints folder, load it with the basic Load Checkpoint node, and use CFG 1 and 4 steps. Wan 2.1 LoRAs generally work, and so do “low noise” Wan 2.2 LoRAs, but not “high noise” ones.Source: Hugging Face model card ·
  • The card lists many versions: a base, V2 to V10, and “MEGA” versions v1 to v12. MEGA is one model for text-to-video, image-to-video and first-frame-to-last-frame, with a workflow in the mega-v3 folder. The card also lists NSFW merge variants; this site does not run any of them and we did not test them.Source: Hugging Face model card ·
  • The author writes: “I do not maintain this anymore and these models are deprecated.” The card says compromises were made for speed and simplicity, so the full Wan 2.2 model pair gives higher quality at longer generation times.Source: Hugging Face model card ·

Claims we did not test

  • Wan 2.1 card: a 5-second 480P video on an RTX 4090 “in about 4 minutes” without optimizations. This is the makers’ own figure; we ran nothing.Source: Hugging Face model card ·
  • Rapid AllInOne card: “Seems to work even on 8GB VRAM”. The author’s informal note, with no test setup, resolution or clip length given.Source: Hugging Face model card ·
  • The Wan team’s paper abstract, quoted in the Diffusers docs, says Wan beats other open and commercial models on its internal and external benchmarks. We do not quote scores: they are the makers’ own, not an independent test.Source: Diffusers docs ·

Not confirmed

  • When Rapid AllInOne was first published: the card gives no date. The commit list shows edits dated Jan 26, Feb 20 and Apr 6, without a year, and an earlier visible commit on Oct 17, 2025, with more history on later pages. We do not guess a release date.Source: Hugging Face commit list ·
  • Resolution and clip length for Rapid AllInOne: the card states neither, so we quote none.Source: Hugging Face model card ·
  • Other repos with the same title: a search showed copies under other accounts, and at least one says its author no longer maintains it. We did not open them, so we cannot say how they differ from Phr00t’s repo. Check the account name before you download.Source: Web search (not Google) ·

Side by side

Wan2.1 T2V 1.3B, Wan 2.2 14B, Rapid AllInOne and this site

Our own table, built from the figures each page states (read 2026-11-19); the makers’ numbers, not our tests. A dash means the page states nothing. The last column is this site’s own settings.
-Wan2.1 T2V 1.3B DiffusersWan 2.2 14B (official)Wan2.2 14B Rapid AllInOneThis site
Made byWan-AI (Wan team)Wan-AI (Team Wan)Hugging Face user Phr00t; base listed as Wan-AI/Wan2.2-I2V-A14BEffect Booth, with a different image model
What it makesVideo from textVideo from text (T2V) or from a photo (I2V)Video; the card lists text, photo and first-to-last-frame modesOne edited photo (1024 × 1024); no video
License (as listed)Apache 2.0Apache 2.0, with a LICENSE.txt that adds use limitsapache-2.0 tagHosted service; see the site terms
Hardware stated8.19 GB VRAMAt least 80 GB VRAM for the single-GPU command“Seems to work even on 8GB VRAM” (informal)None; runs on our servers
Resolution480P recommended; 720P less stable480P and 720P—1024 × 1024
Clip length / speed81 frames at 15 fps (“5-second”); about 4 minutes on an RTX 4090 (claim)81 frames at 16 fps in the Diffusers example—About 37 seconds per photo (measured)
How to runDiffusers WanPipeline, or generate.pygenerate.py, or Diffusers main branch; ComfyUI integrated per the READMEComfyUI Load Checkpoint, CFG 1, 4 stepsUpload at the top of this page
StatusReleased Feb 25, 2025Released Jul 28, 2025Author says it is deprecated and not maintainedUpdated 2026-11-19
Cost to tryOpen weights; your own hardwareOpen weights; your own hardwareOpen weights; your own hardwareFree with sign-up: 10 credits = 1 image; then 10 credits per image, from $0.13 on a monthly plan

Where to run them

How to run Wan2.1 T2V 1.3B Diffusers and Wan 2.2 14B

  1. Wan2.1 T2V 1.3B Diffusers: install Diffusers and load WanPipeline

    Install diffusers, transformers and accelerate, load Wan-AI/Wan2.1-T2V-1.3B-Diffusers with WanPipeline on a CUDA GPU, and ask for 480 x 832 and 81 frames, as the model card does. The card states 8.19 GB VRAM for this model.

  2. Wan 2.2 14B: a big GPU, or offload

    The README’s single-GPU command wants at least 80 GB of VRAM. If memory runs out it lists --offload_model True, --convert_model_dtype and --t5_cpu. The smaller TI2V-5B model lists 24 GB. For Diffusers you need the main branch.

  3. Rapid AllInOne: ComfyUI Load Checkpoint

    Per the card, put the file in ComfyUI’s checkpoints folder, load it with Load Checkpoint, and use CFG 1 with 4 steps. For the MEGA versions follow the workflow in the mega-v3 folder. The author says the repo is deprecated, so check for a newer route before you rely on it.

  4. Need a still image, not a clip? Use this site

    This site currently delivers images only, not video. You can prepare a clean still here, for example a tidier background or even light, and then use it as the input to an image-to-video model somewhere else.

Prompts

Photo prompts for a clean first frame (this site’s editor)

These prompts are for this site’s photo editor, which edits one still photo with a different model from Wan. They do not make video, and we did not run them for this page, so results may vary. Write your own request in the editor too, up to 400 characters.

Clean background for a first frame

Another open video model: LongCat Video

Not run on this site’s editor. Wan 2.2 I2V takes a photo as its input, so a clean still is a sensible start; this site only prepares the photo, it makes no video.

  • Clean, simple background

    Keep my face, hair and clothes exactly the same. Clean up the background so it is simple and uncluttered with soft, even light, and keep a sharp, centred half-length portrait with some space around the head. No text, logos or watermarks.

Even light on the face

Video model news: Seedance 2.5

Not run. Even light is a general good-photo rule, not something the Wan pages say.

  • Even light on the face

    Keep my face and pose exactly the same. Fix the lighting so my face is evenly lit, not too dark or too bright, with natural colours and a calm, softly blurred background. A clean, sharp still photo. No text, logos or watermarks.

Not run for this page. It is the same idea as the one-tap “Sharpen and clean up” edit.

  • Sharpen and reduce noise

    Keep everything exactly as it is. Gently sharpen the photo, reduce noise and compression artifacts, and recover fine detail in hair and fabric without a plastic or over-smoothed look. No text, logos or watermarks.

Pricing on this site

Credit packs or a monthly plan.

Credits come in one-time packs ($4.99 for 50 credits, $9.99 for 120 credits, $19.99 for 300 credits); each image costs 10 credits. Or pick a plan from $9.90 a month (500 credits), cancel anytime.

  • Basic · $9.90/month500 credits a month · about 50 images · $0.20 per image
  • Pro · $19.90/month1500 credits a month · about 150 images · $0.13 per image

Billed every month and renews automatically. Cancel anytime in Settings; you keep the plan until the end of the month you paid for. Each month's credits are for that month; unused ones expire.

Refund policy

FAQ

Questions, answered

What is Wan2.2 14B Rapid AllInOne?

Wan2.2 14B Rapid AllInOne (also searched as “wan 2.2 rapid aio”) is a set of video-model checkpoints published by the Hugging Face user Phr00t, not by Wan-AI. The card calls them mixtures of Wan 2.2 and other Wan-like models and accelerators, with CLIP and the VAE included, in FP8. The author says it is deprecated and no longer maintained.

What is Wan2.1 T2V 1.3B Diffusers?

Wan2.1 T2V 1.3B Diffusers (the repo is Wan-AI/Wan2.1-T2V-1.3B-Diffusers) is the official text-to-video checkpoint from the Wan team, in the Diffusers format, released Feb 25, 2025 under Apache 2.0. Its card states 8.19 GB of VRAM and recommends 480P.

How much VRAM does Wan 2.1 T2V 1.3B need?

The model card says 8.19 GB, and that it can make a 5-second 480P video on an RTX 4090 in about 4 minutes without optimizations. Those are the makers’ numbers and we did not test them. Wan 2.2 14B is bigger: its README says at least 80 GB for the single-GPU command.

Can I run Wan2.2 14B Rapid AllInOne in ComfyUI?

The card says to load the single file with ComfyUI’s Load Checkpoint node and use CFG 1 and 4 steps, and it adds that it “seems to work even on 8GB VRAM”. That last line is informal and untested by us, and the card gives no resolution or clip length.

Is the Wan 2.2 Rapid AllInOne model official?

No. It sits on a personal account, Phr00t, and lists Wan-AI/Wan2.2-I2V-A14B as its base. The official Wan 2.2 models are under Wan-AI, with a GitHub README at Wan-Video/Wan2.2. Copies of the Rapid AllInOne title also exist under other accounts, so check who published the file.

Can I make video with Wan here?

No. This site currently delivers images only, not video, and it does not run Wan or any video model. It edits one photo you upload with a different model. For video, run a Wan model yourself or use a host. Our pages on LongCat Video, Seedance 2.5 and MiniMax H3 Image cover related models.

Which should I pick, Wan 2.1 1.3B or Wan 2.2 14B?

By the stated figures, the 1.3B model needs 8.19 GB of VRAM and targets 480P, while the 14B Wan 2.2 models support 480P and 720P and need at least 80 GB for the single-GPU command. We ran neither, so we cannot rank their quality.

Can I try something here for free?

Partly: new accounts get 10 free credits after signing in, enough for 1 free image (once per account, valid for 7 days, while the day's new-account gifts last). Each image costs 10 credits, from $0.13 per image on a monthly plan.

Need a clean still photo?

Upload it, write the change, and download the result in about 37 seconds.

Edit a photo here

Independent site. Not affiliated with Wan-AI, Phr00t, Hugging Face, GitHub or any host named here, and this site does not run Wan or make video. Wan facts come from the pages named above, read on 2026-11-19; hardware needs, licenses and availability may change.