How AI clone works in IdClone, from one recording to every video you need

Open Clone studio in your account and follow the on screen script. You talk to the camera of your phone or webcam for about two minutes, reading natural sentences and pausing where the guide asks. Then you record a 60 second voice sample in a quiet room. That is all the raw material your clone needs, and you never have to repeat it unless you want a new look.

The quality of this recording sets the quality of every future video, so the studio gives you live feedback while you record: whether your face is centered, whether the light is even, whether the audio is clipping. If something is off, you retake that part before you submit.

Before any clone exists, IdClone confirms that the person in the training footage is the person asking for the clone. You take a short live selfie check in the browser, then read a consent statement out loud that includes a one time phrase generated for that session. The live face and the voice reading the phrase are matched against the training recording. If they do not match, no clone is built.

No identity documents are uploaded, scanned or processed at any point. The check is about you agreeing on camera to a clone of yourself, not about collecting paperwork. The consent recording is stored as the consent record for that clone, and Business and Enterprise accounts can export it for their legal team.

When a team builds a clone of a colleague, the colleague receives their own invite, records their own training video and reads their own consent statement from their own device. An admin cannot upload someone else's footage and create a clone from it. You can read the full rules on our consent and safety page.

After the check passes, IdClone trains two models from your recording: a video clone that reproduces your face, expressions and lip movement in your own framing, and a voice clone that reads any text in your voice. Your account shows the clone as ready, with a short test clip so you can confirm the likeness before you make anything with it.

Your recordings are used only to build and run your own clone. They are never used to train shared models, and nobody else in another account can select your clone. You can delete a clone at any time, and its model files are removed.

Script to video works like a teleprompter that also edits. Paste a full script or write it in place. The editor splits the text into scenes, usually one to three sentences each, and shows the estimated length of every scene and the total, so you see how many plan minutes the video will use before you render anything.

The pronunciation dictionary on Pro and higher plans keeps product names, acronyms and surnames right every time. Add a word once, set how it should sound, and every future script uses it. You can also add pauses, emphasis and scene breaks by hand when the automatic split does not match the rhythm you want.

For each video you pick a framing (head and shoulders, waist up, or side by side with slides), a background (your original room, a solid brand color, an image or a blurred office) and a pace. Pace changes the speaking speed and the length of pauses between scenes, which matters more than people expect: a training update reads well slower, a social clip faster.

With the brand kit you set your logo, colors, intro and outro and caption style once, and every new video starts from it. Templates save a full layout, so a weekly update or a sales follow-up always looks the same.

The storyboard shows every scene with its caption, its timing and the voice preview. Listen to tricky lines, fix a word, move a scene, then check the total length again. On Business and Enterprise plans a reviewer can approve the video before it is rendered or published, and the approval is written to the audit log.

Pick the languages you need and IdClone translates the script, reads it with your voice clone and matches the lip movement in your video clone. Starter covers 10 languages, Pro and higher cover 30+. You see each translated script before rendering and can correct terms, names or tone, which is the step most teams care about for training and legal content. More detail sits on the AI video translator page.

Every translated version is its own video, so its length counts against your monthly video minutes. A three minute training update in four languages uses about twelve minutes.

When you render, you choose whether to show a visible AI-generated label on the video. Content credentials metadata is attached to every export either way, so anyone who checks the file can see it was made with AI. Starter exports in 1080p, Pro and higher in 1080p and 4K. Finished videos download as MP4, voice tracks as WAV or MP3.