AI models on Craze
Every video, image, voice, music and sound model Craze can direct, with what each one makes and what it costs.
Starts at$0.053 per second of source video
ByteDance's motion transfer model that copies movement, expressions, and lip sync from video to bring still images to life.
Craze runs DreamActor 2 when you ask for it by name.
Earlier motion transfer models required extracting a skeleton or pose data as an intermediate step.
Because DreamActor 2 doesn't extract pose skeletons, it generalizes well across different visual styles.
Micro-expressions, hand tremor, the way weight shifts on the feet.
Transfer motion to scenes with multiple characters interacting with objects and each other.










Craze creatively jams, researches and plans with you.
Craze generates anything imaginable. Video, audio, music, motion graphics and more.
Craze watches your footage and builds a final cut. With a real editor if you need it.
Every model here runs on every Craze plan, including Free, and Craze can use a different one for each shot.
| DreamActor 2ByteDance | Seedance 2.5ByteDance Seed Team | Seedance 2.0ByteDance Seed Team | Seedance 2.0 FastByteDance Seed Team | |
|---|---|---|---|---|
| Makes | Text + Image + Video to video | Text + Image + Video + Audio to video | Text + Image + Video + Audio to video | Text + Image + Video + Audio to video |
| Quality | — | Up to 1080p | Up to 4K | Up to 720p |
| Length | — | Up to 30s per clip | Up to 15s per clip | Up to 15s per clip |
| Sound | — | Generated with the picture | Generated with the picture | Generated with the picture |
| From | $0.053 per second of source video | $0.114 per second of video | $0.075 per second of video | $0.06 per second of video |
| Task | Inputs | Resolution | Aspect ratios | Sound |
|---|---|---|---|---|
| Motion transfer | Image, Video | Follows the source | Follows the input | None |
$0/ month
$19/ month
$59/ month
$199/ month
DreamActor 2 solves a specific but important problem: you have great movement but the wrong person. A performer's motion language (the way they gesture, move their head, carry their body) can be transferred to a different character or person. That opens up powerful workflows for animation, reshoots, and content adaptation.
Earlier motion transfer models required extracting a skeleton or pose data as an intermediate step. DreamActor 2 skips that. It reads motion directly from video and transfers it. That means cleaner results and fewer failure modes.
Because DreamActor 2 doesn't extract pose skeletons, it generalizes well across different visual styles. The same motion source animates humans, cartoons, animals, 3D renders, and illustrated characters with equivalent quality.
Micro-expressions, hand tremor, the way weight shifts on the feet. DreamActor 2 captures those subtleties. The transferred motion feels alive, not mechanical.
Transfer motion to scenes with multiple characters interacting with objects and each other. Human-object interaction stays coherent across the motion transfer.
Complicated hand-object interaction, dances, sports moves, and technical gestures transfer cleanly. The model understands spatial relationships and three-dimensional movement.
DreamActor 2 is a transform model: provide a source video with motion and a target image or character, and Craze transfers the movement.
DreamActor 2 works well when both the source motion video and target image have clear, visible subjects. Crowded scenes or footage where the performer is distant may need cropping or focusing first.
The skip-pose-extraction approach sets DreamActor 2 apart. Traditional motion transfer needs an intermediate skeleton or pose data, which introduces failure modes and limits compatibility across visual styles. By reading motion directly from video, DreamActor 2 preserves subtle details that pose-based methods lose. That's why animated characters look as natural as real people, and why hand-object interactions transfer cleanly without needing special handling.
DreamActor 2 is priced by the length of your source motion video at $0.053 per second of source video. A 15-second motion clip costs less than a 90-second performance. Pricing scales linearly with the duration of the motion source. Calculate exact costs on this page.
The target image or character doesn't affect pricing. Only the duration of the source motion video matters.
DreamActor 2 is motion transfer, not avatar generation. Unlike Kling Avatar 2 Pro, OmniHuman 1.5, and VEED Fabric 1.0, which animate a static image to match audio, DreamActor 2 copies movement from one video to animate a different subject.
Character animators and riggers can capture realistic human motion from video and transfer it to their 3D characters or illustrations. No manual keyframing required for natural movement.
Filmmakers can capture a stunt performer's motion and transfer it to an actor's likeness. Safety performers can do the dangerous moves while the lead actor's image carries the motion.
Adapt illustrated characters from manga or comics into animated video using human motion sources. The character design stays true to the source material while movement feels alive.
Choreographers can capture dance from one performer and have it performed by different dancers or characters. The motion transfers, the performer changes.
Filmmakers who can't cast trained performers for complex movement can capture motion from professionals and apply it to their cast. Stunt coordination and movement coaching become less necessary.
Editors can modify performances in post-production by transferring motion between takes or from archived performances. A bad take's motion can be salvaged by transferring it to better footage.
Game developers and VR creators can transfer complex human motion to game avatars and NPCs, creating characters with realistic, nuanced movement.
Streamers and performance artists can use DreamActor 2 to control multiple characters in separate generations from the same performance. One motion source animates different avatars, characters, or stylized renderings in real time, opening up new possibilities for interactive content and virtual events.
Describe what you want the way you would brief an editor, and Craze plans the shots and writes the DreamActor 2 prompts for you. Every Craze plan, including Free, runs DreamActor 2 and every other media model, and Craze can use a different one for each job. Each video lands on an editable timeline with your cuts, captions and music, ready to export as MP4. Generations come out of your plan's monthly credits, from $0.053 per second of source video, with top-ups when you need more.
Animation and character rigging: Character animators and riggers can capture realistic human motion from video and transfer it to their 3D characters or illustrations. Stunt and sports coordination: Filmmakers can capture a stunt performer's motion and transfer it to an actor's likeness. Manga and comic adaptation: Adapt illustrated characters from manga or comics into animated video using human motion sources. Dance and performance video: Choreographers can capture dance from one performer and have it performed by different dancers or characters.
DreamActor 2 is a motion transfer model from ByteDance that copies movement from one video and applies it to a still image. It transfers facial expressions, lip movements, hand gestures, and full-body motion without extracting intermediate skeleton data.
DreamActor 2 costs $0.053 per second of source video on Craze. Longer motion videos cost more. Use the cost calculator on this page for exact estimates.
every Craze plan, including Free include DreamActor 2. Your plan includes monthly credits for motion transfer generations.
DreamActor 2 works on humans, cartoon characters, animals, 3D renders, and illustrated art without special processing or conversion steps.
DreamActor 2 transfers facial expressions, head movement, lip sync, hand gestures, body posture, and complex movement patterns from a motion source to a target image or character.
DreamActor 2 works on humans, cartoon characters, animals, 3D renders, and illustrated art. The model generalizes well across different visual styles.
Craze accepts 1 to 30 seconds of source motion video per generation with DreamActor 2.
Yes. DreamActor 2 can transfer motion between different characters and even handle multi-person scenes with human-object interaction.
ByteDance Seed Team
From $0.114 per second of video
ByteDance Seed Team
From $0.075 per second of video
ByteDance Seed Team
From $0.06 per second of video