AI Lip Sync – Make Any Photo Talk

AI lip sync (also written lipsync) makes a face in a photo speak along with an audio track. The model listens to the speech, predicts the mouth shape for each sound, and animates the lips, jaw, and head so the still image becomes a talking video. On Krea you give it one photo and either a script or a recording, and Hedra or VEED Fabric renders the clip.

Upload a portrait at the top of this page and click Lip sync this photo. The photo opens in the lip sync tool, where you type a script, upload audio, or record your voice, then generate. It takes a few minutes and needs a paid plan.

The best AI lip sync tool is the one that lets you compare models on your own photo, because results vary by face and voice. Krea runs Hedra, the quickest and cheapest per second at 720p, and VEED Fabric 1.0 and Fabric 1.0 Fast at 480p or 720p, side by side. You can also create the photo, the voice, and follow-up video in the same account.

ChatGPT can write the script, but it doesn't lip sync your own photo to a voice. Draft the lines in ChatGPT, then paste them into Krea's lip sync tool: Krea turns the text into speech with an ElevenLabs voice and syncs it to your photo.

No, lip sync needs a paid Krea plan. Free accounts get 100 compute units a day for image generation and other free tools, but lip sync models aren't included, and a 10-second Hedra clip costs about 278 units on its own. The Basic plan's 5,000 monthly units cover roughly 3 minutes of Hedra lip sync; see Krea pricing.

Use a sharp, well-lit, front-facing photo with one person and a clearly visible mouth, as a JPG, PNG, or WebP. For audio, upload a WAV, MP3, or OGG file with clean speech and no music or background noise. Faces turned far to the side, covered mouths, and group photos give weaker results.

Yes. Upload a WAV, MP3, or OGG recording or record your voice directly in the lip sync tool. If you'd rather not record, type up to 2,000 characters and pick one of 21 ElevenLabs voices.

The video is as long as your audio. Hedra uses up to the first 60 seconds of an MP3 or WAV file. Krea doesn't trim audio for the Fabric models, which bill per second, so longer clips cost more.

Hedra renders at 720p. VEED Fabric 1.0 and Fabric 1.0 Fast render at 480p or 720p, which you choose in the lip sync tool; 720p costs more per second. A sharp source photo matters more for the result than the output resolution.

Lip sync follows the sound of the audio, so it works with speech in any language you upload or record. For generated speech, ElevenLabs reads your script in the language you write it in; the voice tool lists English, Spanish, French, German, Italian, Portuguese, Hindi, Arabic, Japanese, Mandarin, Korean, Indonesian. To translate an existing clip, the audio tool dubs video or audio into 25 languages.

Krea's lip sync models start from a still image, so to re-voice someone from a video, use a clear frame from it as the photo. To translate the speech in a video, the audio tool dubs it into another language in the original voices, but it doesn't change the lip movement. To put a different face into a video instead, try face swap.

Most lipsync clips finish in a few minutes. Hedra is usually the fastest and Fabric 1.0 the slowest, and longer audio or busy times take longer. You can leave the page; the video stays in your lip sync session.

Yes, paid Krea plans include a commercial license, so you can use the videos in ads, client work, and products. Only animate faces and voices you own or have permission to use, and label synthetic media where your platform requires it.

No. Krea doesn't add a watermark to the lip sync videos you generate, so you can download and post them as they are.

Your photo and audio are uploaded to your Krea account and sent to the model provider you pick, Hedra or VEED, to generate the video. Krea checks uploads against its content policy before generating. Read the privacy policy for how Krea stores and uses your data.

For making a photo talk, yes. Krea turns one image plus a script or recording into a lip-synced video with Hedra or VEED Fabric, and lets you create the photo, voice, and follow-up video in the same place. Krea doesn't offer stock presenter avatars or translation that re-syncs the lips in existing footage, so for those workflows a dedicated avatar tool may fit better.

Lip sync models focus on the face and head. For full scenes with camera moves, actions, and dialogue, several models in the AI video generator, such as Veo 3.1, generate speech and sound together with the motion.