An AI lip sync generator turns one photo and one voice recording into a video of that person saying it, with the mouth shaped to every word. On Imagera this lives in Lipsync Studio. Upload the photo and the audio, pick a mode, and the recording sets the length of the clip. The audio in the finished video is the file you uploaded, not a generated voice reading your script. This tutorial covers the steps, which mode fits which video, how long a take can be, and how to get a photo and recording that sync cleanly.
Real output from Imagera Hermes Lipsync, September 19, 2026: one portrait and a 9-second voice track, rendered at 768P. The presenter is an invented person, generated for the sample.
1.How to make a photo talk, step by step
- Record the audio first. Speak the line into your phone or a microphone, or export a voiceover you already have. Its length becomes the length of the video, so there is no duration to set.
- Pick the photo. One clear photo of the person who should say it, with the face turned toward the lens and the mouth unobstructed.
- Open Lipsync Studio and choose a mode. Make Any Photo Speak is the default for a single speaker talking to camera. The other modes are in the table below.
- Add the photo and the recording, then set the size. On Imagera Hermes Lipsync you pick a resolution and the clip keeps your photo's shape. On the Fast and Studio passes you pick a frame.
- Check the credits on the Generate button and run it. The finished clip goes to your library, where you can download it again at any time.
2.Which mode to use
Lipsync Studio has seven modes. Each opens its own page with its own uploads:
| Mode | What it makes | What you upload |
|---|---|---|
| Make Any Photo Speak | One person talking to camera | A photo and a recording |
| Same Face, Every Take | A series where every clip is the same person | One reference photo, reused on each take |
| Singing & Performance | A performance with head and body motion, not just a moving mouth | A photo and a song or vocal |
| Two-Character Dialogue | Two people talking in one shot, both in frame the whole time | One photo with both people, and the exchange as one file |
| Music-Reactive Scene | A still turned into a moving scene, with your track as the soundtrack | A photo and a track |
| A Face For Any Voice | A speaker and a room built from the voice alone | Only the recording, plus an optional description |
| Re-Voice Any Video | Your existing footage, with the mouth re-cut to a new voice track | A video you shot and the new audio |
Re-Voice is the one to use when the footage already exists and the words changed. For translating a whole clip into another language, the AI video dubbing tool translates the speech and dubs it.
3.How long can the audio be?
It depends on the engine the mode runs on:
- Imagera Hermes Lipsync, the default on the first four modes above: 5 to about 15 seconds of audio per take, with the clip keeping your photo's shape, at 480P, 768P, 1080P or 2K.
- The Fast and Studio passes, the only engines on Music-Reactive Scene and A Face For Any Voice, and a choice on the other photo modes: 2 to 20 seconds of audio. You pick the frame: Wide (1280 × 704), Tall (704 × 1280), Square (1024 × 1024) or Large (1920 × 1088). Fast is quicker and costs less; Studio spends longer and holds more detail in the face.
- Re-Voice Any Video works on footage you already have, so its limit comes from the clip rather than the recording. It is billed by the second.
A recording that is too long is refused when you add it, with the length named, before any credits are spent. For a longer script, split it into takes at natural pauses. Keep the same photo for every take (or use Same Face, Every Take) and join the clips in your editor.
Frame from a real Imagera Hermes Lipsync output, September 19, 2026: a 13-second music-video take rendered at 1080P. The singer is an invented person, generated for the sample.
4.Getting a photo and a recording that sync well
Most bad lip sync comes from the inputs. Check these before you spend credits:
- Face toward the lens. A three-quarter turn works; a full profile gives the mouth too little to work with.
- Nothing over the mouth. Hands, a microphone, a scarf or heavy shadow across the lips all show up as smearing.
- Even light on the face. Window light or a ring light in front of the person works well. Strong backlight leaves the mouth in shadow.
- A sharp photo. When shooting on a phone, use the rear camera at its normal photo mode rather than a heavy portrait or beauty filter. Filters soften exactly the lip edges the engine follows.
- A clean recording. Record in a quiet, soft room with the phone close to your mouth. Background music or echo on a speech track makes the mouth chase sounds that aren't words.
- A photo shaped like the video you want. On Imagera Hermes Lipsync the clip keeps the photo's shape, so use a tall photo for Reels, TikTok and Shorts. On that engine the photo's shape must fall between 2:5 and 5:2, and the file must be 10 MB or less.
5.Making a trending lip sync video
For a short-form trend, use a tall photo so the clip comes back vertical, and pick the mode that matches the sound:
- Speech, a meme line or a voiceover: Make Any Photo Speak.
- A song: Singing & Performance, which keeps the head and body moving rather than holding a still face.
- A two-person bit: Two-Character Dialogue, with both people in one photo.
Trending sounds are usually someone else's audio. Use your own voice, or a track you have the rights to, so the clip can be posted anywhere. On paid Imagera plans the results can be used commercially under Imagera's terms.
6.What it costs
Lip sync runs are paid in credits. The charge follows the length of your audio and the size you render at: the resolution on Imagera Hermes Lipsync, or the frame on the Fast and Studio passes. The exact figure is on the Generate button before you commit, and a run that fails is not charged.
The Starter plan is $19.99 a month for 240 credits, and credit packs start at 350 credits for $39.99. Every plan is on Pricing. Test the take at a small size first, then render the one you keep at a larger size.
7.What happened to Talking Avatar?
Imagera's older Talking Avatar and Avatar Generator tools were retired into Lipsync Studio, and their links now redirect there. The job is the same: a face, a voice, a clip. Lipsync Studio adds more modes and returns your own audio in the video. Anything you made with the old tools is still in your library.
8.Use faces and voices you have the right to use
Use your own face and voice, invented characters, or people who agreed. Imagera's terms prohibit impersonation, non-consensual likenesses and deceptive content. If you publish in the EU, Article 50 of the AI Act requires a business or professional that uses AI to make a deep fake image, audio or video to disclose that it was artificially generated or manipulated. For evidently artistic, satirical or fictional work, a disclosure that doesn't spoil the work is enough. Under Article 113, the Regulation's general application date is August 2, 2026.
9.Make your first clip
Open Lipsync Studio with one photo and a recording of 20 seconds or less, choose Make Any Photo Speak, and check the credits on the Generate button before you run it.
