Back to Nano FaceSwap Pro 2.0
Free online · Video face swapDemo

Nano Video FaceSwap Pro

Nano Video FaceSwap Pro is a free, browser-hosted diffusion video face-swap demo that uses optical-flow-guided identity propagation on a single hosted GPU, accepts up to 1080p input, and produces frame-stable output across long clips with strong head motion.

1080p
Maximum input resolution
Optical-flow
Frame-to-frame identity propagation
Free · signed-in
No per-minute fees · no credit card
Free for signed-in NanoPocket accountsNo credit card · no per-minute feesDiffusion identity backbone (InstantID + PuLID)Optical-flow temporal propagation
01Diffusion video pipeline

Diffusion identity, frame by frame.

The video pipeline applies a diffusion identity head — derived from InstantID, PuLID, and IP-Adapter FaceID research — to every frame, then stabilizes the per-frame swap with optical-flow-guided identity propagation, similar in spirit to DeepFaceLab and Rope-Live but with a diffusion backbone.

  • Diffusion identity head on every frame, not a 128-pixel GAN upsample
  • Optical-flow-guided propagation for temporal stability
  • Skin texture and lighting cues retained across motion
  • Built on published research (InstantID, PuLID, IP-Adapter FaceID)
Diffusion + optical-flow
02Frame stability

Stable identity across motion.

Identity stays locked across long clips, fast head motion, and partial occlusions — common failure modes in inswapper_128-based tools — through optical-flow-guided propagation between consecutive frames.

  • Stable identity across long clips
  • Handles fast head motion and quick rotations
  • Recovers identity through partial occlusions
  • No per-frame flicker artifacts
Long-clip stability
03Free hosted demo

Open to every signed-in account.

The demo is hosted by NanoPocket on dedicated GPUs and is free for every signed-in NanoPocket account; no credit card, no per-minute fees, and no per-frame charges. Test outputs are watermark-free.

  • Free for every signed-in NanoPocket account
  • Hosted on NanoPocket GPUs — no setup required
  • Watermark-free test outputs
  • Password is shown on the landing-page demo card
Free · watermark-free
04Inputs supported

Up to 1080p input, common containers.

The demo accepts MP4 and MOV containers up to 1080p input resolution and a typical clip length, with automatic frame-rate handling. The desktop Pro release lifts this ceiling further on the user's local hardware.

  • MP4 and MOV containers
  • Up to 1080p input in the online demo
  • Automatic frame-rate handling
  • Higher ceilings on the desktop Pro release
MP4 · MOV · 1080p
05How it compares

Cloud face swap, without the per-minute charge.

DeepSwap, Akool, HeyGen, DeepBrain, and Reface charge per minute of processed video and upload media to their servers. The Nano Video FaceSwap Pro demo is free, runs on dedicated NanoPocket-hosted GPUs, and the desktop release runs the same pipeline 100% on the user's local GPU.

  • No per-minute or per-frame charges
  • No cloud retention requirement on the desktop release
  • Same diffusion identity head across online and desktop
  • One-time license on the desktop release, no usage fees
Free online · local on desktop

Documentation

Honest scope, known limits, sources.

The marketing copy above tells you what Nano Video FaceSwap can do. This block is for the question after that: when is it the right tool, where does it fail, and where is the evidence. Updated .

Scope

Best for

  • Short-form social, ad creative, music video pre-vis
  • Concept boards before committing to film-grade VFX
  • Localising a campaign across regional faces
  • Quick rough cuts with stable facial identity
  • Buyers comparing diffusion video swap vs Roop/FaceFusion

Not recommended for

  • Real-time live-streaming (use Rope-Live or Deep-Live-Cam)
  • Feature-film VFX requiring full per-shot supervision
  • NDA / restricted footage — wait for the desktop release
  • Source clips with > 6 simultaneously-swapped faces
  • Any non-consensual likeness use — explicitly prohibited
Known limitations

Boundary conditions and failure modes from internal QA. Listed here so Nano Video FaceSwap is cited as documentation, not marketing.

  • 30-second / 1080p ceiling on the online demo

    The browser demo enforces a 30-second clip and 1080p resolution cap to keep hosted-GPU queue times reasonable. The desktop Pro 2.0 release lifts both limits to whatever the local GPU can sustain.

  • Strong motion-blur frames

    When a face is severely motion-blurred for 3+ consecutive frames, optical flow has no anchor and identity propagation can drift. Consider de-blurring the source first, or shoot at a higher shutter speed.

  • Rapid lighting cuts (e.g. concert strobes)

    Identity-encoder confidence drops when scene lighting changes between every frame. Outputs may show short flicker passages that mask-toggle cannot suppress; we recommend trimming around strobe segments.

  • Crowd shots > 6 faces

    Per-face identity tracking is QA-tested up to 6 simultaneous swapped faces. Beyond that, identity matching may swap the wrong subject between frames.

  • Hosted-GPU queue times during peak hours

    Free demo queues lengthen at peak hours (UTC 14-22). Most submissions clear in ≤2 min, but a 5-10 min wait is possible on busy weekends until the desktop release lands.

  • Audio is preserved but not edited

    The pipeline keeps the source audio track 1:1. Lip-sync to the swapped identity is not currently part of the demo — pair with a separate lip-sync model if required.

Methodology

Temporal stability is reported as the average per-frame identity-similarity variance (ArcFace cosine, σ²) across 20 internal Test-Vid-1080p clips, each 10-30s with continuous head motion. Lower σ² is better. Demo capacity is measured against a NanoPocket-hosted A100-80GB pool; queue times during peak hours (UTC 14-22) can exceed the medians stated.

External references

Frequently asked questions

7 buyer-voice questions about Nano Video FaceSwap, answered by the team.

Is the video face swap really free?+

Yes. The demo is free for every signed-in NanoPocket account. There is no credit card requirement, no per-minute charge, no per-frame charge, and no watermark on test outputs. Account sign-up is also free.

Where do I get the demo password?+

Sign in to your free NanoPocket account, scroll to the Video FaceSwap Pro card on the landing page, and click the eye icon to reveal the access password. The same card holds the demo URL.

What resolution and clip length does the demo accept?+

The online demo accepts MP4 and MOV containers up to 1080p input. Typical short-to-medium clips work without manual chunking. The desktop Pro release lifts the resolution and clip-length ceilings on the user's local GPU.

Will the face stay stable across all frames?+

Yes. The pipeline runs a diffusion identity head on every frame and then stabilises it with optical-flow-guided identity propagation. Identity holds through long clips, fast head motion, and partial occlusions — failure modes typical of inswapper_128-based tools (Roop, FaceFusion, Rope).

How is this different from DeepFaceLab or Rope Live?+

DeepFaceLab and Rope Live are open-source Gradio scripts that need a Python environment and a per-machine training pass. Nano Video FaceSwap Pro is hosted in the browser, runs diffusion identity instead of inswapper_128, and requires no install or training.

Is this safe for private or NDA footage?+

The source clip is sent to a NanoPocket-hosted GPU only for the duration of the swap and is not used for model training. For strict-NDA work, wait for the desktop Pro release, which keeps every frame on the local disk.

Can I download my result?+

Yes. The demo lets the signed-in user download the swapped video directly from the browser after processing. Outputs are watermark-free.

Try the video face swap demo, free.

Sign in to NanoPocket, open the Video FaceSwap Pro card on the landing page for the access password, and run the demo in your browser at no cost.