PAN2: what it is, what the ComfyUI files are, how to run it

PAN2 usually means PAN 2, a 29B-parameter open video world model from MBZUAI that makes clips of up to 20 seconds from text and an optional first image, with Comfy-Org's repack for ComfyUI; this site currently delivers images only, not video, but its photo editor, a different model, 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

  • Not PAN 2, not affiliated with MBZUAI or Comfy-Org
  • Images only: this site does not make video
  • Free with sign-up: 10 credits = 1 image
parameters of the diffusion model, per the original model card
29B
licence of PAN 2’s own code and weights
Apache-2.0
longest clip in the model card; 704 × 1248 at 24 fps
20 s
the int8 model file in the ComfyUI repack (bf16: 67.37 GB)
26.68 GB

What PAN2 is

From the original model card (IFM/PAN2)

  • PAN 2 is developed by the Institute of Foundation Models (IFM) at MBZUAI. The card calls it a general-purpose world model: given a current observation (an optional image plus text) and actions written in text, it simulates the outcome as a video of up to 20 seconds.Source: Hugging Face model card (IFM/PAN2) ·
  • Design, as the card describes it: a language model writes the next world state as a structured description, and a 29B-parameter diffusion transformer trained from scratch turns it into video latents. The text encoder is Qwen3.5-9B and the Wan 2.2 VAE decodes the result.Source: Hugging Face model card (IFM/PAN2) ·
  • Output and defaults in the card: text-to-video or image-to-video from one checkpoint, 704 × 1248 at 24 fps, 50 sampling steps, guidance 2.5, 241 frames (about 10 seconds) by default and up to 481 frames (20 seconds).Source: Hugging Face model card (IFM/PAN2) ·
  • Licence: PAN 2’s own code and weights are Apache 2.0. Wan 2.2, Qwen3.5-9B and Diffusers keep their own licences. The card lists out-of-scope uses, among them impersonating real people, deepfakes, misinformation and sexual content involving minors, and strongly recommends running it behind the separate pan2-guardrail.Source: Hugging Face model card (IFM/PAN2) ·
  • Dates: the original repo was created on 2026-10-07 and last changed on 2026-11-12 (Hugging Face API); the card’s citation says October 2026. The Comfy-Org repack was created on 2026-11-15.Source: Hugging Face API ·

The makers’ claims and limits, not tested by us

  • The card reports capability tables, public benchmarks and a human preference study from its own technical report. We do not quote the scores; none is an independent test, and we did not read the report.Source: Hugging Face model card (IFM/PAN2) ·
  • The card’s own limits section says about two in three simulated multi-step plans contain at least one step that is not carried out, and that the model is sensitive to prompt wording. Free-text prompts should be rewritten into a structured world-state caption by a language model first.Source: Hugging Face model card (IFM/PAN2) ·

The ComfyUI repack (Comfy-Org/PAN2)

  • The card is short: “Repackaged model files for ComfyUI”, a link to IFM/PAN2 as the original, tags for text-to-video and Apache 2.0, and the folders to put the files in. It gives no workflow, no steps, no author names and no date.Source: Hugging Face model card (Comfy-Org/PAN2) ·
  • Files in the repo: two diffusion models (bf16 and int8), two text encoders (a Qwen3.5-9B bf16 and an int8 version) and the Wan 2.2 VAE. The card’s folder list names only the int8 text encoder, but the repo also holds the bf16 one. Sizes are in the table below.Source: Hugging Face file list (Comfy-Org/PAN2) ·
  • A community news site, ComfyUI Wiki, which says it is separate from official ComfyUI, reported on November 16 that PAN 2 loads in ComfyUI core after pull request 17755 merged on November 15, that it recommends the int8 file, that no official workflow template exists yet, and that one-line prompts give noise. Its RTX 5090 timing for 241 frames was 15.2 seconds per step for int8 and 26.8 for the dense build. We have not run any of this.Source: ComfyUI Wiki (third party) ·

Other things called PAN2

  • In SuperCollider, an audio programming language, Pan2 is a two-channel equal-power panner. It is a different thing.Source: SuperCollider docs ·
  • Most search results for the bare word are about PAN2, the human gene for a poly(A)-specific ribonuclease subunit, from gene databases and antibody sellers. We saw those only as search-result titles, because the pages blocked our fetcher, and they have nothing to do with the video model.Source: Web search results (title only) ·

On this site

Not confirmed

  • A GPU or memory minimum: neither the original card nor the repack card states one. The original card says each worker holds the full 29B model and names a recent software stack (Python 3.14, torch 2.12.1, FlashAttention 4).Source: Hugging Face model card (IFM/PAN2) ·
  • Who at Comfy-Org made the repack, and whether MBZUAI endorses it: the card says neither. The ifm-ai GitHub page credits the PAN Team, and it names the technical report without linking it.Source: GitHub (ifm-ai/PAN2) ·
  • Whether the repack can be used for image-to-video: its card is tagged text-to-video only, while the original supports both and the third-party article describes an image-to-video node. We have not tested either.Source: Hugging Face model card (Comfy-Org/PAN2) ·

Sizes

PAN2 file sizes in the ComfyUI repack, and this site

Our own sums from the Hugging Face file list of Comfy-Org/PAN2, read on 2026-11-19, in decimal GB. These are download sizes, not memory needs; no card states a memory figure. This site’s numbers come from its own settings.
-Size on Hugging FaceWhat it is
pan2_bf16.safetensors67.37 GBThe full-precision diffusion model
pan2_int8_convrot.safetensors26.68 GBThe int8 diffusion model
qwen3.5_9b_bf16.safetensors19.31 GBText encoder, full precision (in the repo, not in the card’s folder list)
qwen3.5_9b_int8_convrot.safetensors9.75 GBText encoder, int8 (named in the card’s folder list)
wan2.2_vae.safetensors1.41 GBThe video decoder
Int8 route in total (our sum)37.84 GBint8 model + int8 text encoder + VAE
Full-precision route in total (our sum)88.09 GBbf16 model + bf16 text encoder + VAE
This siteNothing to downloadOne photo edited in about 37 seconds, 10 credits (1024 × 1024); no video

Where to run it

How to run PAN2 in ComfyUI or on your own GPU

  1. Know which repo you are downloading

    IFM/PAN2 is the original from MBZUAI, in the Diffusers layout. Comfy-Org/PAN2 is the single-file repack for ComfyUI. The repack card links back to the original and adds no instructions.

  2. In ComfyUI: put the files in the model folders

    Per the repack card: the diffusion model in models/diffusion_models, the text encoder in models/text_encoders and the VAE in models/vae. A third-party article says you need a ComfyUI build newer than v0.45.0 and a hand-built graph, since no official workflow template exists yet.

  3. Write structured prompts

    The original card says PAN 2 works best with structured world-state captions, so rewrite a free-text idea with a language model first. A one-line prompt may give poor results.

  4. Without ComfyUI: the original repo

    The GitHub and Hugging Face pages give an inference script (generate.py) and an environment with Python 3.14 and recent torch. They state no GPU minimum, so check before you rent hardware.

  5. Need a still image instead? Use this site

    This site currently delivers images only, not video. Upload one photo and describe the change in the editor, up to 400 characters.

Prompts

Photo edit prompts for this site’s editor

These prompts are for this site’s photo editor, which edits still images with a different model from PAN 2. We did not run them for this page, and they do not make video.

One prompt per one-tap edit

  • Remove background clutter

    Edit my uploaded photo: remove distracting clutter from the background (stray objects, litter, bins, cables, signs and people in the background) and fill those areas naturally with what is around them. Keep the framing as it is. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Plain studio background

    Edit my uploaded photo: replace the background with a plain, softly lit light-grey studio backdrop with natural soft shadows. Keep the framing as it is. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Professional headshot

    Edit my uploaded photo into a clean professional headshot: a smooth neutral grey-blue background with a gentle gradient, soft even studio light on the face, the same crop or a slightly tighter head-and-shoulders crop. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Fix the lighting

    Edit my uploaded photo: fix the lighting only. Balance the exposure, lift dark shadows on faces, tame blown-out highlights and correct the white balance so skin tones look natural. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Warm golden-hour light

    Edit my uploaded photo: change only the light to warm, low golden-hour sunlight from one side, with soft long shadows and a warm glow, as if taken shortly before sunset. Keep the scene and framing. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Classic black and white

    Edit my uploaded photo into a classic black and white photograph with rich contrast, deep blacks, clean whites and fine film grain. Keep the composition exactly as it is. Keep every face, expression, pose and outfit exactly the same. No text, logos or watermarks.

  • Watercolor painting

    Repaint my uploaded photo as a soft watercolor painting on textured white paper, with loose washes and visible brush edges, keeping the same composition. Keep every person and pet recognizable, and do not add or remove anyone. No text, logos or watermarks.

  • Sharpen and clean up

    Edit my uploaded photo: gently sharpen it, reduce noise and compression artifacts, and recover fine detail in hair, fabric and edges, without a plastic or over-smoothed look. Keep every face, expression, pose and outfit exactly the same. 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 PAN2?

In AI video, PAN2 means PAN 2, an open world model from the Institute of Foundation Models at MBZUAI. From text and an optional first image it simulates an action as a video of up to 20 seconds, using a 29B-parameter diffusion model, and its own code and weights are Apache 2.0.

What is the difference between Comfy-Org PAN2 and IFM PAN2?

IFM/PAN2 is the original release in the Diffusers layout. Comfy-Org/PAN2 is a repack of the same model as single files for ComfyUI, in bf16 and int8, with a text encoder and the Wan 2.2 VAE. Its card names no authors and adds no usage steps beyond folder placement.

Is this site PAN2, and can I make video here?

No. This site is not PAN 2, MBZUAI or Comfy-Org and does not run PAN 2. It currently delivers images only, not video. Its own editor uses a different model: upload one photo and write the change (up to 400 characters), or tap a ready-made edit.

How do I run PAN2 in ComfyUI, and is there a workflow?

Put the repack files in ComfyUI’s diffusion_models, text_encoders and vae folders, as the Comfy-Org card shows. A community news site reported that ComfyUI core supports it since November 15 and that no official workflow template exists yet, so the graph is built by hand. We have not run it.

How much VRAM does PAN2 need?

Not stated by either model card. What we can give is download size, summed by us from the file list on 2026-11-19: 37.84 GB for the int8 route (model, text encoder and VAE) and 88.09 GB for the full-precision route. Download size is not the same as memory use.

Is PAN2 free, and can I use it commercially?

The weights are free to download. PAN 2’s own code and weights are Apache 2.0, while Wan 2.2 and Qwen3.5-9B keep their own licences, so check those too. The card also lists out-of-scope uses, such as deepfakes and impersonation. This is not legal advice.

What else is called PAN2?

PAN2 is also the human gene for a poly(A)-specific ribonuclease subunit, which fills most search results for the bare word, and Pan2 is a stereo panner in the SuperCollider audio language. Neither has anything to do with the video model.

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.

Just need one photo edited?

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

Edit a photo here

Independent site. Not affiliated with MBZUAI, the Institute of Foundation Models, Comfy-Org, ComfyUI or Hugging Face, and this site does not run PAN 2 or make video. Facts about the model come from the makers’ own pages and the third-party pages named above, read on 2026-11-19; the makers’ claims are not tested by us. This is not legal advice.