Frame continuity
Watch the face through turns, blinks, speech, and camera movement. Look for identity jumps or sudden shape changes.
How to face swap a video online
How to Face Swap a Video Online: Start With the Hardest Frames
Use a five-step browser workflow: measure the clip locally, test its hardest movement, match a clear reference face, confirm the exact credit total, and review the whole result before sharing.
Current product walkthrough
See the upload, output, and pricing controls before you start
This silent 40-second walkthrough was captured on July 26, 2026 and may show the retired 720p selector. The current service now fixes output to the highest 1080p tier. It shows navigation and controls only; it does not present a generated result or imply a quality benchmark.
- 0–8 seconds: page and limits. The recording shows the older five-minute and dual-tier interface; current limits are 10 minutes and fixed 1080p.
- 8–16 seconds: upload controls. Choose Video Deep Swap and locate the source-video and target-face upload slots.
- 16–24 seconds: output options. Review face-only processing and the visible controls without uploading private media.
- 24–40 seconds: preparation and cost. Use the preparation guide, then return to the workspace to read the exact pre-submission credit amount.
Clip scouting
Find the frames most likely to break continuity
| Clip condition | Risk | Practical response |
|---|---|---|
| Slow, front-facing motion | Lower | Use as the first representative test. |
| Fast head turns or profiles | Higher | Match the reference angle more closely and inspect transition frames. |
| Hands, hair, microphones, or objects cross the face | Higher | Mark the timestamps and judge whether the occlusion is essential. |
| Motion blur or low light | Higher | Use a clearer source clip when possible; upscaling cannot restore missing landmarks. |
| Hard cuts or major lighting changes | Variable | Test each scene separately before processing a long edit. |
Five-step workflow
How to Face Swap a Video Online in Five Steps
- Trim a representative segment. Remove unused lead-in, end frames, and alternate takes while keeping at least one important turn or occlusion.
- Inspect the target frame by frame. Confirm that the face remains large, clear, and visible through the scenes that matter.
- Choose a clear reference image. Use a sharp JPEG, PNG, or WebP reference with a pose close to the most important scene.
- Confirm output and cost. Confirm the fixed highest 1080p tier and read the final credit amount calculated from rounded-up seconds in the workspace.
- Generate and review key timestamps. Inspect turns, blinks, speech, occlusion recovery, edges, and lighting across the entire result.
Do not optimize only for the first frame. The shortest useful test contains the hardest motion in the project, not merely an easy opening pose.
Formats, limits, and cost
Estimate the complete task before submission
Before uploading, use the free local video preflight to read browser-visible file size, declared type, dimensions, aspect ratio, duration, rounded-up billed seconds, and face-only credit estimates without sending the selected file. It does not inspect codec, frame rate, audio, face visibility, identity continuity, or output quality.
The current video workspace accepts MP4, MOV, or WebM target videos and JPEG, PNG, WebP, HEIC, or HEIF face references. A reference image can be up to 30 MB; a target video can be up to 95 MB and 600 seconds, and both files must be 95 MB or smaller combined. The interface calculates the required amount before submission.
Current person-mapping scope: one face reference is applied to the primary on-camera person. The workflow does not assign different reference faces to several people in the same video. For several people in one static group photo, use multiple face swap.
| Processing mode | Output tier | Current rate | Current minimum |
|---|---|---|---|
| Face-only, scene preserved | 1080p highest tier | 1 credit per rounded-up second | 5 credits |
The workflow changes the face only and preserves the original body, clothing, hands, products, text, logos, framing, and scene; it does not accept broader character or scene replacement. The displayed amount before submission remains authoritative.
Result review
Review motion, not just resemblance
Occlusion recovery
Inspect the frames before, during, and after an object crosses the face.
Lighting continuity
Check whether facial shading follows scene light through cuts or exposure changes.
Edge integrity
Review hairline, ears, glasses, hands, and jaw edges at the intended playback size.
For short looping motion, use the GIF face swap guide. Use the image quality guide when the problem is already visible in a single frame, and review the privacy checklist before sharing another person's likeness.
Consent and privacy
Process only faces and videos you have permission to use
Confirm that every real person agrees to the edit and its intended audience before upload. DeepSwapAI does not use uploads to train AI models; uploaded and generated media is removed from DeepSwapAI servers within 24 hours. Review the Privacy Policy before sharing another person's likeness.
How this guide was made
Built from current workflow limits and observable failure conditions
The DeepSwapAI Product Team rechecked the current video MIME types, reference formats, file limits, 600-second duration, fixed 1080p pricing, browser-local preflight boundary, queue, 24-hour deletion flow, and public workspace controls on August 21, 2026. The walkthrough records only the public interface and does not contain customer media or a generated result. The guide does not invent an accuracy score, completion time, or result guarantee; see the verification methodology.
Test the hardest useful segment
Choose a short section that includes motion or occlusion, confirm the displayed credit estimate, and review continuity before processing the full clip.