Trim a representative segment
Remove unused lead-in, end frames, and alternate takes while keeping an important turn or occlusion.
Swap faces in videos up to 10 minutes at 1080p online. No watermark, files auto-deleted in 24h. 3 credits per second; packs from $19.99, no subscription. DeepSwapAI video uploads can be up to 200 MB each — about 2–3 minutes of typical 1080p phone video, so compress or trim longer clips.
Before you pay: check your clip on this device without an account. Video costs 3 credits per second with a 12-credit minimum, and the exact total appears before you submit. See all prices
Added to the workspace below.The workspace is still loading. Please try again in a moment.Try a sample
Pick a sample to compare one real run before and after. Nothing is uploaded and no credits are used.





Everyone shown is an AI-generated demo image. The video and GIF are camera moves over an AI-generated still, not filmed footage.
DeepSwapAI Studio
Upload, map, preview, and switch between photo, batch, group-photo, video, and GIF workflows in one focused workspace.
Photo workflow
Upload a source photo and a face reference, then replace the identity while keeping the scene.
Output
Controls
Free local video preflight
Choose an MP4, MOV, or WebM file to read its size, declared type, dimensions, aspect ratio, duration, billed seconds, and estimated fixed highest 1080p cost.
MP4, MOV, or WebM · up to 200 MB and 600 seconds
Local only: the selected video stays on this device and no file bytes are uploaded.
The browser reads native video metadata locally and shows a preview only after you choose a file.
Select a file to start the local preflight.
Face-only processing with scene preservation uses max(12, rounded-up seconds × 3) at the fixed highest 1080p tier. The final amount shown in the workspace before submission is authoritative.
Real output
Everyone shown is an AI-generated DeepSwapAI demonstration image, not a real person. We show one completed run of this workflow exactly as it came back.
The target clip is a slow camera push-in over an AI-generated still, so it shows the swap on moving frames, not natural motion.

Face photo used · Charged: 12 credits
No customer upload, stock photo or real person was used. Each result comes from one completed run on a paid account (no trial watermark) and is shown without retouching; web copies are resized and re-encoded. One example is not a quality benchmark. See the inputs and run details
Five-step video workflow
Remove unused lead-in, end frames, and alternate takes while keeping an important turn or occlusion.
Confirm that the face remains large, clear, and visible through the scenes that matter.
Use a sharp reference with a pose close to the most important scene.
Confirm the fixed highest 1080p tier and read the final credit amount calculated from rounded-up seconds.
Inspect turns, blinks, speech, occlusion recovery, edges, and lighting across the complete result.
Choose video by motion, not only length
A short difficult turn can matter more than a long easy shot. Test the motion, occlusion, lighting change, and cut most likely to break continuity before processing the complete video.
Open the video preparation guideStart with a clear face at useful size and inspect speech, blinking, profile turns, and any frame where the subject leaves the camera.
Use a representative clip that includes the real motion and lighting range. A static opening frame is not enough to validate the complete edit.
Trim unused footage first, test the hardest short segment, then confirm the fixed highest 1080p credit total before submitting the complete duration.
Mark every occlusion and hard cut. Review the frame before, during, and after each event so a temporary face loss or identity drift is not missed.
The current video workflow replaces only the primary on-camera person with one reference face. It does not map different references to several people in the same video; use multiple face swap for group still photos.
Verified workflow facts
The selected quality, rounded-up duration, minimum charge, and required credit total are shown before submission.
Accepted media
MP4 · MOV · WebM Face: JPEG · PNG · WebP · HEIC · HEIFUpload limits
600 seconds maximum Video up to 200 MB · face image up to 30 MBOutput tier
Highest 1080p fixed No lower-resolution billing tier1080p pricing
3 credits per second Rounded up · 12-credit minimumWorked examples
These examples use the current production formulas, not an invented average.
Applied ruleVideo: max(12, rounded-up seconds × 3 credits) at fixed 1080p.
| Video duration in seconds | Applied rule | Required credits |
|---|---|---|
| 1 | max(12, 1 × 3) | 12 |
| 8 | max(12, 8 × 3) | 24 |
| 12 | max(12, 12 × 3) | 36 |
One-time credit packs
Buy credits once and use them across photo, batch, group-photo, video, and GIF workflows. No recurring charge or automatic renewal.
Basic
600 credits
≈ 200 billed video seconds · 12-credit minimum/video
Choose $19.99 starter packStandard
Best first purchase900 credits
≈ 300 billed video seconds · 12-credit minimum/video
Get 900 creditsPremium
Save 5% per credit3,150 credits
≈ 1,050 billed video seconds · 12-credit minimum/video
Get 3,150 creditsFAQ
Direct answers about browser-based video input, supported formats, duration limits, resolution pricing, and retention.
Yes. Upload one target video and one clear face-reference image in the browser; no desktop software or local GPU is required.
The current workflow accepts MP4, MOV, or WebM video up to 600 seconds and 200 MB. The face-reference image can be up to 30 MB. 200 MB is about 2–3 minutes of typical 1080p phone video; compress or trim longer clips.
Measured duration is rounded up to the next whole second. The fixed highest 1080p tier costs 3 credits per second with a 12-credit minimum. The exact total appears before submission.
Not in the current video workflow. It accepts one face reference and replaces the primary on-camera person only. Use multiple face swap for different people in one static group photo.
DeepSwapAI removes uploads and generated media from its servers within 24 hours. The selected media is sent to an inference provider only to complete the requested task. It is not used to train models.
No. It uses the browser’s native video reader on this device and does not send the selected file to DeepSwapAI or a third party. A file is uploaded only if you separately submit it in the generation workspace.
Setup is free. Credits are only used when you generate.
Practical guides
Video preparation
Troubleshooting
Output evaluation
GIF loops
Private input preflight
Image quality
Privacy and consent
Developer integration