Image to video
Animate an image you own into a 5- or 10-second clip with synchronised audio.
Image to video turns a still image into a 5- or 10-second clip with sound. It needs an active subscription or a free video use, and it only works on images you own. Free video uses come, rarely, as a lucky draw bonus perk; each covers one 5-second clip at no credit cost, and the dialog shows how many you have left. Without a subscription or a free use left, Generate Video opens a preview of the feature instead.
Starting a render
The Generate Video button appears on a gallery image when two things are true: the image is yours, and it doesn't already have a video. You'll find it in two places.
- Under an AI-generated image in a chat, on the row of actions with Edit image and Delete Image.
- On a character's gallery, on any image you own.
You own an image if you created the character it belongs to, or if you generated the image yourself. There is no entry point from moments, and an avatar is only animatable through the gallery.
Each generation costs 120,000 credits per 5 seconds of video; a 10-second clip costs 240,000. The charge lands when the finished video is attached to the gallery item, so a render that fails costs nothing — but Generate Again is a second full-price generation, not a retry of the first.
The AI Video Director
The dialog is a shot planner, not a list of presets.
Design directions, or skip straight to writing
Design 3 directions sends the image to an AI pass that studies it and proposes three distinct shots for the length you picked — each with a title, a subject motion, a camera move, an intensity and a sound direction. Pick one and it fills in the controls; Redesign asks for three more. When the image comes from one of your chats, it also reads the message the image belongs to and the one before it, so a direction can play what happens next in the story.
This is an extra AI call, billed on its own token use and waived while your subscriber allowance covers it. Writing Subject motion yourself skips it entirely.
Direct the motion
Set what moves, how strongly, and what the camera does.
Direct the sound
Choose a sound mode, and — if you want the character to talk — type the exact words.
Generate
The render streams its progress through directing, rendering and uploading.
Motion controls
| Control | Options |
|---|---|
| Subject motion | Free text, up to 1,200 characters |
| Motion intensity | Subtle · Natural · Expressive |
| Camera motion | AI chooses · Locked camera · Slow push in · Slow pull out · Pan left · Pan right · Tilt up · Tilt down · Gentle handheld |
| Output framing | Original · Portrait 3:4 · Landscape 16:9 · Vertical 9:16 · Square 1:1 |
| Optional ending frame | Another image from the same character's gallery |
Picking an ending frame changes how the camera behaves: the start-to-end transition takes over the camera path, so the clip lands on the second image.
Sound
The model generates audio inside the video, synchronised with what happens on screen — ambience, foley, music and character vocal sounds. It is not a soundtrack added afterwards.
| Sound design | What you get |
|---|---|
| AI chooses | The model picks the audio direction from the image and the motion |
| Ambience & foley | Environment and the sounds of what moves |
| Background music | Music under the shot |
| Character voice | The character makes vocal sound — and speaks your dialogue if you write any |
| Silent | No audio |
Four text fields sit under the mode, and each appears only when it can do something:
| Field | Limit | Shown when |
|---|---|---|
| Sound direction | 500 characters | The mode is anything but Silent |
| Exact dialogue (optional) | 500 characters | The mode is AI chooses or Character voice |
| Voice tone | 160 characters | You have typed dialogue |
| Spoken language | 80 characters | You have typed dialogue |
Leave Exact dialogue empty and the model will not invent lines — that is what the placeholder means by avoiding invented dialogue.
What comes out
| Property | Value |
|---|---|
| Duration | About 5.2 seconds (124 frames), or about 10.1 seconds (243 frames) at 10s |
| Frame rate | 24 fps |
| Format | MP4 with an audio track |
| Original framing | Keeps the source aspect at roughly a 1280 × 720 pixel area, rounded to a 64-pixel grid |
| Portrait 3:4 | 832 × 1088 |
| Landscape 16:9 | 1280 × 704 |
| Vertical 9:16 | 704 × 1280 |
| Square 1:1 | 960 × 960 |
Length picks 5s or 10s. A 10-second clip costs twice as much, and a free video use covers 5-second clips only.
Combine pictures
Combine pictures is a second mode in the same dialog, for subscribers only — free video uses cover Animate this image only. It builds one clip from the image plus up to 3 more pictures, and costs 180,000 credits per 5 seconds of video.
Pick the extra pictures under References, one picture per character:
| Tab | What it offers |
|---|---|
| This chat | Characters in the chat you opened the dialog from |
| Recent chats | Characters from your recent chats |
| My characters | Characters you created |
| Favorites | Characters you favorited |
| Personas | Your personas that have a portrait |
| Uploads | Photos you uploaded here or in a scene image, and pictures you recently sent in your chats |
The character you are animating never appears in the list — its image is already the first picture. Upload takes a PNG, JPG, WebP or AVIF image up to 12 MB; each upload is checked before it is saved, and Delete upload removes it for good.
The shot controls change with the mode: Together replaces Action with two-person interactions, from Hold hands and Kiss to explicit ones, and the one-tap scenes are this mode's own. There is no ending frame in this mode. Two people give the most reliable results; with three or more, the model can occasionally mix people up.
The clip opens with the first added person already beside your character: before the render, their picture and your image are combined into one opening picture, posed the way the shot begins. If your motion has them arrive, such as walking in or coming through a door, the clip opens on your image as it is and they enter during it. A third or fourth person always enters during the clip, a picture of an object or a place is never put into the opening, and if combining fails, the clip opens on your image alone.
Adult content
The video step does not add a content filter on top of the image you started from. The dialog says as much in its own pitch: "Unrestricted generation for eligible adult content".
Two limits apply all the same:
- For a character whose creator gave them an age under 18, the video service's own content filter stays on. A clip it blocks isn't delivered and costs nothing.
- In Combine pictures, a picture that looks like a child or a photo of a real-looking person — an upload or a persona portrait — keeps the same filter on, and a picture that clearly shows a recognizable public figure can't be used. When the filter blocks a clip, the message says whether a minor or a real person's photo turned it on.
While it renders
One render at a time, per account. Starting a second while the first is going returns "Another video is still rendering. Wait for it to finish before starting a new one."
Closing the dialog does not cancel anything: "Generation continues in the background. We'll let you know when it's done." While you stay in the app, a message pops up when the video is ready. If you leave or lose the connection first, you get a notification instead — in the bell, and as a push notification if you've turned those on — saying the video is ready or that the render failed. A failed render costs nothing. The server gives up on a render after 10 minutes.
The finished video
The video is attached to that one gallery image. The image action row gains Show video / Show image to flip between them, and anyone who can see the gallery image can play the video. Only you get the generate and delete actions.
Deleting a video is permanent. It removes the clip, deletes the stored file and restores the still image. Nothing is refunded, and animating the image again is a new 120,000-credit render.
Related
- Generating images — where the source image comes from
- Subscriptions — the plan requirement for video
- Credits — where the 120,000 lands
- Snapshots — publishing gallery content to a character's page