Image-to-video
Animate any character image into a short clip. Seven motion presets plus a custom-prompt option. Premium subscription required.
Image-to-video turns a static character image into a short animated clip — a turn of the head, hair in the wind, a smile and blink. The clip saves to the character's gallery and can be shared as a video file. The button appears below any AI-generated character image in chat.
Source post: Image-to-video generation →
What it costs
| Service | Cost per generation |
|---|---|
| Image-to-video (SeeDance I2V) | 120,000 credits |
Video generation costs an order of magnitude more than image generation because GPU time on the underlying model is expensive — we pass that through at cost.
Subscription required. The tool is gated to active subscribers. Generated videos are visible to everyone (free users can view them in chats and galleries), but generating new ones needs a plan.
Where to use it
The video button appears below any AI-generated character image in:
- Chats — animate the character's reply image
- Character profiles — animate avatars
- Moments — animate an image attached to a moment
The motion presets
Seven directed presets plus a custom prompt option:
| Preset | Best for |
|---|---|
| Gentle movement | Calm character moments, ambient motion |
| Turn to camera | A character noticing you |
| Hair in wind | Atmospheric, romantic |
| Smile and blink | Greeting moments, warm interactions |
| Wave hello | Playful, social |
| Breathing | Subtle, immersive |
| Look around | Awareness, exploration |
| Custom | Write your own motion description |
The presets are tuned to produce reliable, in-character motion. Custom descriptions work but require some prompt-writing skill — short, specific phrases work better than narratives.
Custom motion examples
Stick to one or two beats. Long descriptions confuse the model and produce worse animations.
Generation time
1-2 minutes per video. You can keep using Reverie while it runs — the result lands as a notification when ready.
What the videos look like
3-5 seconds, 720p, MP4. Compatible with every social platform. No watermark.
Best results:
- Character portraits with clear faces
- Mid-range shots with visible expression
- Soft backgrounds that allow motion to feel natural
Worst results:
- Full-body action poses — limbs flicker
- Crowded scenes — every other element competes for the motion budget
- Very dark images — motion gets lost in shadow
What you can't do
- Speech / lip-sync. Videos are silent. Mouth animation is generic — don't expect lip-sync to a voice clip.
- More than 5 seconds. Longer clips aren't supported yet.
- Audio. Videos export silent.
- NSFW motion. Image-to-video has a safety layer that's stricter than the image side. Some NSFW images fail to animate.
What's planned for graduation
When image-to-video moves out of Labs (target: 2026), expect:
- Longer clips (up to 10 seconds)
- Voice sync (optional matched lip movement)
- Better cost-per-generation as the underlying model gets cheaper
- Inline in moments and chat by default (currently behind the Labs toggle)
Tips
- Generate from your best image. Bad input = bad video.
- Keep motion subtle for emotional scenes. Tiny movements read as alive.
- Save and reuse. Generated videos persist in your gallery and can be reattached to multiple moments/chats.
- Share as MP4. Generated clips work on every platform — Twitter, TikTok, Instagram, Discord.