How to Create an AI Avatar Video (Free, From One Photo)

To create an AI avatar video you need one photo, one audio clip and about five minutes. Upload the photo of the person who should speak, upload the audio they should speak with, pick a scene preset, and generate. The whole flow runs in the browser on this site and the finished MP4 has no watermark.

Step 1: Pick the right photo

The photo decides how well the face animates. A front-facing shot with one person filling the frame, even light across the face and eyes visible gives the model the most to work with. Waist-up portraits render best because the avatar can move head and shoulders naturally, not just the mouth.

Things that hurt the render: sunglasses, hats shadowing the eyes, mirror selfies, group photos, heavy filters, low light. JPG, PNG and WebP all work, up to 10MB. You must have the rights to the photo; that rule applies to every step here.

Step 2: Record the audio

The audio drives the speech and the video length. An mp3, wav or m4a of up to 15 seconds fits the standard tiers. Record close to the mic in a quiet room; background noise is the main cause of mushy sync. Two to three short sentences is plenty for a first test.

Use speech you own or have permission to use. Do not upload celebrity voices, copyrighted characters or commercial music.

Step 3: Choose scene preset, resolution and length

The studio offers talking presets (Clean Portrait, Podcast Guest, News Desk, Street Interview) that frame the render, plus a performance mode (Dance Clip, Cinematic Scene) that renders a prompt-driven clip instead. Resolution runs 480p to 1080p; length tiers are 5, 10 and 15 seconds, and the final video follows your audio.

Talking renders run at 720p or 1080p and cost 80 and 120 credits at 10 seconds. New accounts start with 10 free credits, which covers a 5-second 480p test in performance mode, enough to check how your photo renders before paying anything.

Step 4: Generate, review, download

Hit generate and the render usually finishes within minutes; the page updates itself. Watch the result once with sound: check that the face reads as the person, that the mouth tracks the words through the whole clip, and that the scene fits the platform you are posting to. Download the MP4 and it is yours, watermark-free, portrait or landscape.

If the sync disappoints, the fix is almost always in the inputs. Retake the photo in better light, re-record the audio cleaner, and run it again. Failed renders refund their credits automatically.

Start now in the studio on the home page.