Turn family photos into video greetings with MiniMax H3
MiniMax H3 is the AI video model from MiniMax: it turns a photo card, or a few lines about the moment, into a 5-10 second clip with sound. It is our Auto pick for video and animates every video scene in the catalogue. MiniMax also makes the Speech model that reads greetings in your own voice.
- Made by MiniMax
- from 380 credits
- 5-10 s
- With sound
Updated 29.09.2026
Candles, a sunset blanket, a toast and a first kiss, all made with H3
Prompt
A kitchen birthday, candles out in one breath, then clapping and a big hug
Prompt
A hilltop, the valley below, the sun going down, both wrapped up tight
Prompt
Dishes everywhere, grandparents at the head, glasses meeting in the middle
Prompt
A flower arch, guests on their feet cheering, petals falling like snow
What Auto means when it picks MiniMax H3
It is what Auto means for video
Leave the Model picker on Auto and your video is made with MiniMax H3. It also animates all 20 video scenes in the catalogue, from "Make a wish, Mum" to "First kiss as newlyweds", so any scene clip you have watched here shows what H3 does with a family photo.
Every clip comes with sound
H3 always adds sound: a breath over the candles, clapping, wind in the grass, glasses meeting over a table. There is no silent mode, and a greeting rarely wants one - a silent birthday looks like security camera footage.
It costs less per second
At 768p, a second of H3 costs a fraction of a second of Veo 3.1 and well under Kling 3.0 Pro. Over a whole clip the difference can pay for another one - say, a version for the grandparents in another city.
It starts from a picture you like
Make a photo card first, and when the faces and the light are right, bring it to life: H3 starts from that exact picture. In a scene, both steps run one after the other for you. With no picture at all, it makes the clip from your words, in 16:9, 9:16 or 1:1.
Short warm moments, yes; singing and 15 seconds, no
H3 is at its best with short, warm moments where one small action lands: candles out in one breath, glasses meeting over the dishes, a kiss under a flower arch, two people under one blanket while the sun goes down. The four clips above were all made with H3 at 768p with sound. The faces and the room come from the card; H3 adds the movement and the noise.
It is not the answer to everything. If people on screen need to sing or say words you can make out, Veo 3.1 generates the voices and singing together with the picture - its page has a family singing the end of Happy Birthday. If you want a clip longer than 10 seconds or no sound at all, Kling 3.0 Pro goes to 15 seconds in 1080p with sound on or off. For a calm clip that plays on repeat, look at Wan 3.0 Prime. And if the budget is the only thing that matters, Veo 3.1 Lite starts lower.
Three to six minutes from a scene to your family moving
The catalogue holds 36 scenes, and 20 of them are video. Each one is a ready moment - a kitchen birthday, a long family table, a wedding toast by the sea - drawn with Nano Banana Pro. You add a photo for each person, we make the card with their faces in it, and MiniMax H3 animates it with the scene's motion and sound. The whole thing takes 3-6 minutes, about as long as writing the message that goes with it.
A few things help. Up to 6 people fit in one picture, though faces stay most reliably like themselves with 4 or fewer, so for a big family choose the four who matter most for this card. A clear, well-lit photo where the face is easy to see does more than a group shot from across the room. And a finished card can be made sharper or reshaped, which is handy when a tall card meant for a phone needs to become a wide one.
The other MiniMax: Speech 02 HD and your own voice
The second MiniMax model here makes no pictures at all. MiniMax Speech 02 HD is behind "your own voice": on a paid plan - Plus, Family or Pro - you record yourself once in the browser, reading aloud one sentence we show on screen, and from then on a voice greeting can be read in your voice instead of a catalogue one. Making the voice is a one-time charge in credits; the table below has the number.
Your voice also works in a talking card. The photo speaks your line - up to 15 seconds of speech - with Kling AI Avatar v2 moving the face and head, or OmniHuman 1.5 moving the whole person. It is the one to send to someone who still keeps your old voicemails. Without a paid plan, the 8 catalogue voices of ElevenLabs Eleven v3 still read your greetings, and you can hear each one before choosing.
Two ways to start an H3 clip, and when to send it
There are two ways in. In Standard, choose Video and leave the Model picker on Auto, which already means H3. Or tell the agent in the chat what you picture and answer its questions one at a time. Either way the price is on the button before you press, a video that fails gives its credits back, and the finished clip goes out by email, SMS or a messenger, straight away or on the morning of the day.
Plan for the wait. A scene takes 3-6 minutes, so make it the evening before and schedule it: "Make a wish, Mum" can land on her phone on the morning of her birthday, and "One blanket, one sunset" on your anniversary, with no weather forecast to check. For a family toast, pick the four faces that matter most for this card, because up to four is where likeness holds best.
Keys, a blanket, a long table: H3 prompts with sounds you can name
A video for my brother Sam's new flat: he stands in the empty living room holding up his keys, morning sun through bare windows, boxes around him. He throws the keys up, catches them and laughs. Sound: keys jingling, an echo in the empty room.
A video for our 10th anniversary: my wife and I on a hilltop at sunset, wrapped in one knitted blanket, the valley glowing below. The grass ripples, she leans her head on my shoulder. Sound: evening wind, rustling grass, distant birds.
A video for Grandpa's 80th: a long table in the evening, dishes and flowers everywhere, Grandma and Grandpa at the head. Everyone raises a glass to Grandpa, he laughs and wipes his eyes. Sound: glasses clinking, family chatter, one big laugh.
H3 by the second, your voice once: today's MiniMax prices
Here are today's numbers for both MiniMax models. When you choose a video model, look at the price per second first: at 768p, H3 costs less per second than Kling 3.0 Pro or Veo 3.1, and it runs every scene. Kling 3.0 Pro earns its price with 15-second clips and a sound switch, Veo 3.1 with voices and singing.
| Model | Makes | Strong at | Length | Quality | Sound | Price |
|---|---|---|---|---|---|---|
| MiniMax H3MiniMax | Video | Best value | 5, 6, 8, 10 s | 768p, 2K | With sound | from 380 credits · 75 cr a second |
| MiniMax Speech 02 HDMiniMax | Voice | Your own voice | Your own voice: 1,880 credits, once |
Prices are in credits and are always shown on the button before you press it. Pricing
Keep exploring
Kling is Kuaishou's family of AI models, and three of them work here: Kling 3.0 Pro for video, Kling O3…
OpenVeoVeo is Google's video model, here in three versions: Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite. Each turns…
OpenAI video greetingAn AI video greeting starts from a photo of your people and ends as a short clip where they move…
OpenAI voice greetingAn AI voice greeting is your written greeting turned into a voice message they can play on speaker, in the…
OpenMiniMax H3 and MiniMax Speech, answered
MiniMax H3 is an AI video model made by MiniMax. It turns a picture, or a few lines of text, into a short clip with sound: 5, 6, 8 or 10 seconds, at 768p or 2K. On Wish Generator it is the Auto choice for video and animates all 20 video scenes in the catalogue.
MiniMax H3 makes clips at 768p or 2K. All four scene clips on this page were made at 768p with sound, and on a phone screen that is plenty. Pick 2K when the clip will play on a TV or a laptop at the party; the button shows the price for whatever you choose. Started from words alone, H3 can make a wide 16:9, a tall 9:16 or a square 1:1 clip.
Yes - every H3 clip has sound, and it cannot be switched off. In our scenes it carries the sounds around a moment: clapping, wind, clinking glasses. If you want people on screen to sing or speak clearly, Veo 3.1 is made for that. For a silent clip, pick Kling 3.0 Pro or Veo 3.1: both have a sound switch, and silence costs less per second there.
Up to 10 seconds: you can choose 5, 6, 8 or 10. That is enough for one action and a reaction - candles out, then the hug. For a longer take, Kling 3.0 Pro goes up to 15 seconds. A long message sits better in a song or a voice greeting.
Yes. Add JPG, PNG or WebP photos up to 10 MB from your device, your library or People, your saved contacts. In a scene, the card with their faces is made first and H3 animates it. A picture can hold up to 6 people; faces stay most reliably like themselves up to 4.
Yes, if you are on a paid plan: Plus, Family or Pro. MiniMax Speech 02 HD makes your own voice after you read one sentence aloud in the browser. Voice greetings and talking cards can then be read in your voice. Making the voice is a one-time charge in credits, shown in the table above.
Twenty video scenes are waiting for your people
Pick a scene, add their photos, and MiniMax H3 brings it to life while you write the words that go with it.
Create with MiniMax H3
