How do I make a photo talk with AI?
Upload a clear, front-facing photo to a talking-photo tool such as HeyGen's photo avatars[1], Hedra[4] or Kling's Avatar 2.0[6], then give it a script or a voice recording, and the model animates the mouth, head and expression to match the speech. Free plans work for short tests but usually add a watermark and cap the length.
What you need before you start
- A photo with a readable face. Front-facing or a slight three-quarter angle, mouth visible and closed or relaxed, even light, nothing covering the lips. Sunglasses, hands at the chin and heavy motion blur all make the mouth harder to animate.
- A voice. Either type a script and pick a synthetic voice, or upload your own recording. Kling's Avatar 2.0, for example, takes the avatar image (which is also the first frame), a speech audio file and an optional text prompt[7].
- The right to use the face. Use your own photo, a character you made, or a photo you have permission for. TikTok does not allow AI-generated content showing the likeness of anyone under 18, or of adult private figures without their permission[9].
- A target shape. Crop the photo to the format you will post (9:16 for Shorts, Reels and TikTok) before you upload, so the tool does not invent the missing edges.
Step by step
- Pick the tool by what you need most: language coverage, length or price. HeyGen markets photo avatars that speak over 175 languages and dialects[1]; Hedra describes its Character 3 model as audio-to-video for talking and singing clips[4].
- Upload the photo. Check the face detection preview if the tool shows one; if it boxes the wrong area, re-crop tighter.
- Add the voice: paste the script and choose a voice, or upload audio. Keep the first test under 15 seconds so a bad result costs little.
- If the tool accepts a prompt, describe the delivery, not the words: "calm, small nods, smiles at the end".
- Render, watch it at full size, and fix the photo or the audio rather than re-rolling the same inputs. Most lip-sync errors come from a blocked mouth or noisy audio.
- Download. On free plans, expect a watermark: HeyGen says all its videos carry a logo watermark by default and removing it needs a paid plan[3].
How long can a talking photo be?
| Tool | Stated length |
|---|---|
| HeyGen Free plan [2] | 3 videos a month, each up to 1 minute |
| Hedra Free plan [5] | Capped at 20 seconds, watermarked (per Hedra's in-app plan copy) [5] |
| Hedra Character 3 [4] | Up to 10 minutes of talking or singing video |
| Kling Avatar 2.0 [6] | Claims full coverage for 5-minute content scenes |
| D-ID [8] | Each video is rounded up to the nearest 15 seconds when minutes are deducted |
Talking photo or a character who talks in a scene?
A talking photo keeps the original framing: a head and shoulders that speak. If you want the character to walk, react and speak inside a scene, you want a video model with native audio instead. In our own Seedance renders, a line written in quotes in the prompt comes out spoken with matching mouth movement, with no audio upload. The trade-off is that you do not control the exact voice. The page on how AI lip sync works explains both methods, and talking avatar or character video helps you choose.
If the photo is a drawing rather than a person, see can AI animate a drawing. Character.AI describes its AvatarFX as bringing a pre-existing image to life to speak, sing and emote[10], which is one route for illustrated characters; we could not verify who can use AvatarFX today or its limits.
Where Cast fits
- You do not have a photo and want a character who speaks a short line inside a 5 to 15 second scene, voiced by the video model.
- You want to describe a new character in a sentence and pick from four drawn looks (15 credits) instead of finding a usable photo.
- You want to try it on your phone first: the first 5-second video is free with no sign-up.
Where it does not:
- Making an existing photo talk. Cast cannot take an uploaded photo or drawing; characters are drawn from a text description.
- Lip sync to your own recorded audio. Cast's voices come from the video model.
- Talking-head clips longer than 15 seconds or a digital avatar of yourself.
People also ask
- Do I need to record my own audio?
- Not usually. Most talking-photo tools accept a typed script and a synthetic voice, while some avatar models expect a speech file: Kling's Avatar 2.0 takes speech audio alongside the image[7].
- Is there a free way to make a photo talk?
- Yes, with limits. HeyGen's Free plan includes 3 videos a month up to 1 minute each[2], and Hedra's in-app plan copy lists a Free plan capped at 20 seconds with a watermark[5]. See the best free talking photo AI for a fuller comparison.
Sources
Claims about other companies come from their own pages, read on the date shown. Products and prices change; check the link for the current version.
- HeyGen: photo avatar (read )
- HeyGen: pricing (read )
- HeyGen help: how to remove the HeyGen watermark (read )
- Hedra: models (read )
- Hedra: plans (read )
- Kling AI: Avatar 2.0 user guide (read )
- Kling AI: Avatar 2.0 user guide (inputs) (read )
- D-ID: Studio pricing (read )
- TikTok: AI-generated content FAQ (read )
- Character.AI: AvatarFX announcement (read )
- Character.AI: AvatarFX safety measures (read )
Related answers
- What is the best free talking photo AI?
- How does lip sync work in AI video generators?
- Can AI animate a drawing of my character?
- Do I need a talking avatar or an AI character video?
- Can I make an AI avatar of myself?
- Is it legal to make an AI video of a real person?
Part of the Cast answers library (Talking photos and avatars).
Updated