On this page
To make an AI video from a photo, upload a sharp, well-lit image to an image-to-video tool, pick 9:16 for TikTok, Reels or Shorts, and write a prompt that describes only the motion plus what must not change. Generate, review for face or logo drift, then add captions and music before exporting an MP4.
The model is rarely what decides the result. In Revid's study of AI videos vs real footage, AI-generated TikToks averaged twice the views of real footage, while on X, Instagram and Threads AI visuals clustered in the lowest engagement tier. Same technology, opposite outcomes: the photo you start from, the prompt and the edit around the motion are what separate a clip people watch from one they skip.
How do I turn a photo into an AI video?
The short version, using Revid's Image to Video Effect tool:
- Upload one high-quality JPG or PNG with a clear subject.
- Choose 9:16 vertical for TikTok, Reels or Shorts (16:9 for YouTube or a website).
- Describe the motion, for example: "camera slowly pushes in, hair moves gently in the wind, warm sunset light, keep the face natural".
- Generate the clip and watch it at full speed.
- Regenerate with a simpler prompt if the face, hands or logo drift.
- Add a hook caption, music and a call to action in the editor.
- Export an MP4 and post or schedule it.
The process is the same whether you call it photo to video, image to video or picture to video. What changes the result is which workflow you pick, so start there.
Which kind of AI photo video are you making?
Most disappointing results come from using the wrong workflow. Ten wedding photos in a single-image animator give you one moving photo, not a highlight reel. A portrait in a montage tool gives you a slideshow, not a character.
| What you want | Workflow | Revid tool |
|---|---|---|
| One still image that moves | Image to video | Image to Video Effect or AI Image Animator |
| Many photos in one video | Photo montage | Photo to Video Converter |
| Photos told as a story with voiceover | Narrated story | Media to Story |
| A person or mascot doing a dance or gesture | Motion transfer | Motion Transfer |
| A face that speaks a script | Talking avatar | AI Talking Avatar |
Quick rules for choosing:
- One photo, one short clip: use the Image to Video Effect. You can describe the motion or let the AI choose. The AI Image Animator gives you more control over duration, AI voice and aspect ratio.
- Several photos: the Photo to Video Converter arranges them with transitions, music and pacing. Switch to Media to Story when the photos need a narrated thread, such as a trip, a founder story or a case study.
- Movement from a reference video: Motion Transfer takes one motion video and one photo, and applies the dance or gestures to the person or character in the photo. Only animate people who agreed to it.
- A talking photo: you need a portrait, a script or audio, and a voice. The AI Talking Avatar builds the presenter video with lip sync, captions and music.
How to choose a photo that animates well
The model doesn't move your pixels around. It reads the subject, depth and perspective, predicts motion and synthesizes new frames from what it can see. A clearer input means fewer guesses, and fewer guesses mean fewer warped faces.
| Use a photo with | Avoid |
|---|---|
| One clear subject, separated from the background | Busy, cluttered backgrounds |
| Even lighting that shows depth | Deep shadows over the face |
| High resolution and sharp edges | Blur, especially on hands and faces |
| The whole head and subject in frame | Cropped heads, subjects touching the frame edge |
| No text near the edges | Text or tiny logos baked into the image |
| A natural lens | Distorted wide-angle selfies, overlapping people |
If the photo shows a real person, use only images you have the right to use (more on consent below).
Step by step: make an AI video from a photo in Revid
Step 1: Pick the workflow for your output
Decide the output and the platform before you open a tool, using the table above. If you have one photo and aren't sure, start with the Image to Video Effect.
Step 2: Upload your photo
Upload the original file, not a screenshot or a copy forwarded through a messaging app, which has already lost detail. For a montage or a story, upload all the photos you want the AI to work with.
Step 3: Choose the aspect ratio before you generate
The ratio shapes how the AI frames the subject and where the motion happens, so pick it first. Cropping a horizontal clip to vertical afterwards cuts off the subject and softens the image.
| Ratio | Use for |
|---|---|
| 9:16 vertical | TikTok, Instagram Reels, YouTube Shorts, Stories |
| 1:1 square | Square feed posts, repurposed social content |
| 4:5 | Tall Instagram feed posts |
| 16:9 horizontal | YouTube, websites, presentations |
Step 4: Write a motion prompt
This step decides most results. The photo already holds the subject, setting, colors and light, so the prompt should describe movement and limits, not the scene. Use this formula:
[Camera motion] + [Subject motion] + [Environment motion] + [Lighting or mood] + [Preservation rule]
Strong: "The woman smiles subtly and turns her head slightly toward the camera. Her hair moves gently in a soft breeze. The camera slowly pushes in. Warm golden-hour light. Keep her face natural and preserve her identity."
Weak: "Make this photo cinematic and viral." It leaves every decision to the model.
| Weak prompt | Better prompt |
|---|---|
| Make this cool | Camera slowly pushes in, background lights flicker, subject stays still and sharp |
| Make it cinematic | Slow dolly-in, shallow depth of field, warm sunset light, subtle wind movement |
| Animate this product | Smooth studio camera arc, moving reflections, product shape unchanged |
| Make her smile | Subtle natural smile, slight head turn, gentle blink, preserve facial identity |
One of each motion type is enough. Two camera moves compete, and three environment moves turn chaotic. When in doubt, add a constraint instead of another action.
Step 5: Generate, then iterate
Revid runs several AI video models under one subscription (Seedance 2.5, MiniMax Hailuo H3, Wan 3.0 and Google Veo 3) and lets you choose the model per scene. Treat the first generation as a read on what the model understood; the second or third is usually the keeper. Starting from a photo keeps those rounds short: across 713,649 clips generated on Revid, starting from an image halved the wait on the base tier, 18 seconds instead of 39.
Step 6: Review in three passes
Watch the clip at full speed, not scrubbed, three times:
- Identity. Does the subject still look like the photo? For people: face, age, hair, eyes, expression, proportions. For products: shape, logo, label, color, packaging.
- Motion. Look for morphing hands, warped features, melting backgrounds, floating objects, odd blinking and flickering edges.
- Feed. Does the first second stop a scroll, is the subject visible on a phone, and is there room for captions?
Fix identity drift with a preservation rule ("keep the product shape, logo, label and color exactly the same"). Fix odd motion by asking for one subtle movement: "The person remains in place, smiles subtly, and the camera slowly pushes in."
Step 7: Edit the clip into a post
The generated clip is your moving visual layer, like b-roll. Captions, music and a hook turn it into a post. Many viewers start with the sound off, so captions carry the message. In Revid's timeline editor you can add synced captions (30 styles, 100+ languages), music, an AI voice-over from 1,000+ voices in 70+ languages, and text overlays.
A simple 8-second structure:
| Time | What happens |
|---|---|
| 0 to 1.5 s | Hook text appears immediately |
| 1.5 to 4 s | The main AI motion plays |
| 4 to 6.5 s | Benefit, reveal or story context |
| 6.5 to 8 s | Call to action, or end on the opening frame so it loops |
Caption lines you can adapt:
- Product: "One product photo. Three ad variations. No reshoot."
- Memory: "A photo from 1987. Brought back to life."
- Reveal: "This was just one photo."
Prompt to try
Make an 8-second vertical 9:16 product video of a white leather sneaker on a clean studio background. Open with the hook text "New drop. One photo. Full video." in the first second, then a slow camera arc with soft reflections moving across the shoe. Premium, minimal style with light electronic music, ending on the opening frame so it loops.
Step 8: Export in the right format
For short-form feeds, export a 9:16 MP4. TikTok's Creator Academy recommends 9:16 at 1080x1920 pixels, and Instagram accepts Reels from 1.91:1 to 9:16 at a minimum of 30 FPS and 720 pixels. For YouTube, square or vertical videos up to three minutes count as Shorts; YouTube's Shorts help page covers the upload side. Revid exports in 720p, 1080p or 4K at 30 or 60 fps, so choose 1080p or higher for phone screens. Keep faces and text clear of the app buttons with the free TikTok safe zone checker.
Step 9: Post, test and make more versions
Revid publishes or schedules directly to TikTok, Instagram and YouTube, and each post reports its views, likes and comments, so you can see which version won. One photo can feed several tests: a slow premium push-in, a fast zoom meme cut, a voiced version, a caption-only version for sound-off viewing, and a before-and-after that shows the original photo first.
Motion prompt templates you can copy
Paste these into the Image to Video Effect or the AI Image Animator and adjust.
| Photo type | Prompt | Best for |
|---|---|---|
| Portrait | The person slowly turns their head toward the camera and gives a subtle natural smile. Eyes blink gently. Hair moves slightly in a soft breeze. Camera slowly pushes in. Keep the face, age, clothing and identity unchanged. | Profile photos, creator intros |
| Product | The camera moves around the product in a smooth studio arc. Subtle reflections move across the surface. Clean background. Do not alter the logo, label, shape or packaging. | E-commerce, ads, launches |
| Landscape | Slow cinematic push-in. Clouds drift gently from left to right. Water ripples naturally. Leaves move in the wind. No new buildings or people. | Travel, tourism |
| Food | Steam rises gently from the food. Camera slowly pushes in. Background stays softly blurred. Keep the dish unchanged. | Restaurants, cafés |
| Old family photo | Very subtle motion only: gentle blink, slight breathing, soft push-in. Preserve the original face, clothing, hairstyle and photo style. Do not modernize the image. | Tributes, anniversaries |
| Real estate | Slow dolly forward into the room. Curtains move slightly. Sunlight shifts across the floor. Keep the architecture, furniture, layout and proportions unchanged. | Listings, rentals |
| Illustration | Animate while preserving the illustration style. Subtle parallax depth. Light particles drift slowly. Do not change the character design, colors or line art. | Artists, album art |
| Fashion | The model holds the pose while fabric moves in a light breeze. Camera slowly tilts up. Keep the face, outfit, pattern and accessories unchanged. | Lookbooks, brands |
| Pet | The dog blinks, tilts its head slightly and breathes naturally. Fur moves gently. Keep the face, color, markings and body shape unchanged. No cartoon effect. | Pet accounts, gifts |
| Meme | Sudden dramatic zoom. Background shakes slightly. Dust and light rays move behind the subject. Keep the subject recognizable and centered. | Reaction edits, gaming |
Motion words to build your own prompts
- Camera: slow push-in, dolly forward or back, pan, tilt, orbit, handheld micro-shake, parallax depth, rack focus.
- Subject: blink gently, slight smile, subtle head turn, hair moves in breeze, breathing motion, model holds pose.
- Environment: clouds drift, water ripples, steam rises, smoke swirls, rain falls, neon flickers.
- Constraints: keep identity unchanged, preserve product shape, keep logo readable, no morphing, no extra people, stable background, natural motion only.
Why is my AI video from a photo blurry or full of artifacts?
Blur and artifacts almost always trace back to three causes: a low-resolution or compressed source, more motion than the model can keep consistent, or a clip stretched longer than the model holds up. Fix the cause, not the symptom.
How to preserve the original resolution
- Upload the original camera file, ideally at least 1080 pixels wide. A screenshot or a chat-app copy has lost detail the model can't recover.
- Generate in the final aspect ratio. Cropping or resizing afterwards scales the video up and softens it.
- Export at 1080p or 4K, and upload that MP4 directly instead of screen-recording it or re-saving it through another app.
- If edges still look soft, regenerate the scene with another model; models differ in how much fine detail they keep.
How to improve frame consistency
- Ask for one subject motion and one camera motion at most.
- Add a preservation rule for whatever must not change (face, logo, room layout).
- Keep single-photo clips around 5 to 10 seconds, and build longer videos from several short scenes.
- Keep hands still or out of frame, and prefer camera movement over body movement.
| Problem | Likely cause | Fix |
|---|---|---|
| The face changes | Too much facial motion, weak source photo | "Subtle blink and slight smile only. Preserve the original face, age, hairstyle and identity." |
| Hands distort | Hands are hard for video models in motion | "Keep the hands still. Only the camera slowly pushes in." |
| The logo turns unreadable | Small text distorts in motion | "Keep the logo and label unchanged. Use only slow camera movement." Then add the logo as an overlay. |
| The background melts | Too much freedom or a busy background | "Keep the background stable. Only the clouds move gently." |
| Motion is too boring | Prompt too subtle | "Slow cinematic push-in, subtle parallax between subject and background." |
| Motion is too chaotic | Too many actions in one prompt | "Smooth slow push-in. Subject remains still. Natural motion only." |
| The subject is cut off in the feed | Wrong ratio or outside the safe zone | Generate in 9:16 from the start and keep the subject centered. |
Why does my AI anime portrait look distorted or uncanny?
Anime portraits break for the same reasons, plus one more: the model drifts toward realistic skin and lighting. Light the source evenly from the front, use a front-facing or three-quarter pose with the face unobstructed and the hands out of frame, and name the style in the prompt ("preserve the anime line art and flat cel shading, no realistic skin texture"). Keep the motion small: a blink, some hair movement and a slow push-in. Our guide to making an AI anime video from a photo has full prompts.
AI photo video examples by use case
| Input | Workflow | Prompt or instruction | Edit and export |
|---|---|---|---|
| A sneaker photo | Single image | Slow camera arc, subtle reflections, keep the logo, sole, color and laces unchanged | "New drop. One photo. Full video." in 9:16 |
| A pasta dish | Single image | Steam rises, slow push-in, warm restaurant light, dish unchanged | "Tonight's special is already calling." in 9:16 for Reels and Stories |
| 20 wedding photos | Photo montage | Warm romantic pacing, soft transitions, piano music, short captions | Couple names and date as overlays, in 9:16 and 16:9 |
| A living room photo | Single image | Slow dolly forward, sunlight across the floor, layout unchanged | "3-bed apartment. Morning light." in 9:16 for social, 16:9 for the listing page |
| 12 travel photos | Narrated story | A warm weekend-in-Lisbon story with a light voiceover and captions | 9:16 |
| A mascot image plus a dance clip | Motion transfer | Apply the dance while keeping the character recognizable and centered | 9:16 |
Checklist before you post
- Photo: sharp subject, good light, nothing important cut off, no tiny text.
- Prompt: specific motion, one camera move, one mood, preservation rules.
- Video: identity preserved, no warped hands or logos, believable motion, a clean start and end.
- Social: a hook in the first second, captions readable on a phone, subject centered, text inside the safe zone, the right ratio, a visible call to action.
Is it legal to animate someone's photo?
It depends on rights, consent and context, so make consent your default. Before animating a real person, check that you own the photo or have permission to use it, that the person agreed to be animated (take extra care with children), and that the result doesn't make them appear to say, do or endorse something they didn't. Copyrighted characters, artwork and brand logos need rights too. Revid's Terms of Service make users responsible for what they create and prohibit infringing anyone's rights.
Platforms also ask you to label realistic AI content:
- TikTok requires a label on AI-generated content with realistic images, audio or video, may apply one automatically to uploads carrying C2PA Content Credentials, and says the label doesn't affect distribution for videos that follow its Community Guidelines.
- YouTube requires disclosure of realistic AI-generated or altered content through the "AI use" setting in YouTube Studio, and since May 2026 labels undisclosed photorealistic AI automatically. Disclosure alone doesn't limit reach or earnings.
- Instagram and Facebook require Meta's AI disclosure for photorealistic video that was digitally created or altered, and show the label as "AI info".
TikTok holds its own tools to the same standard: videos made with TikTok's AI Alive photo-to-video feature, launched in May 2025, are labeled as AI-generated and carry C2PA metadata.
FAQ
Can AI turn a photo into a video?
Yes. Image-to-video models generate new frames from a still image to simulate camera and subject motion. Sharp, well-lit photos with one clear subject, plus prompts that describe specific movement, give the cleanest results.
How long should an AI video from a photo be?
Aim for 5 to 10 seconds for a single animated photo and 15 to 45 seconds for a montage or narrated story. YouTube treats square or vertical videos up to three minutes as Shorts, but short clips are easier to watch through and to test.
Can I use AI photo videos for ads?
Yes, if you have the rights to the image, product, music, voice and any person shown. TikTok's ad format policy requires ad video to be dynamic and says static images should not make up more than half of the ad, so animate the photo rather than running it still.
Can I animate old family photos?
Yes. Use very subtle motion (a gentle blink, slight breathing, a soft push-in) and tell the model to preserve the original face, clothing and photo style so it doesn't modernize the person.
What's the difference between the Image to Video Effect and the AI Image Animator?
Both start from one still image. The Image to Video Effect is the most direct path from photo to animated clip, while the AI Image Animator adds more control over duration, AI voice and aspect ratio.
Is Revid free for making videos from photos?
Sign-up is free, no credit card needed, and new accounts get free credits to try video creation. Regular creation then uses the monthly credits of a paid plan, starting at $39 a month. You can estimate a video's credit cost before generating it.
