DreamActor 2
DreamActor 2 is a video model by ByteDance. In Vini it has one operation, Avatar (Аватар): the picture you connect as the first frame decides who appears, the background and the framing; the source clip decides the body movement, facial expression and lip movement. It does not write a scene from text and does not edit the finished clip. To replace the people in an existing clip while keeping its background and framing, use Replace hero instead.
When to choose it
Section titled “When to choose it”The author describes real people, animation and pets as inputs. Choose DreamActor when the hero is a cartoon, an animal, a puppet or several characters in one picture, and a person acts the performance in the clip. It also works when the people in the source clip are cut off at the knee or waist, where the Avatar address of Kling refuses a clip without a complete upper body.
What goes in and what comes out
Section titled “What goes in and what comes out”Connect one picture as the first frame and one source clip of up to 30 seconds. The studio refuses a longer clip before any charge. There is no prompt field at this address: text, negative prompts and character names are not sent, and the studio says so on the node instead of dropping them silently. Direct the result by choosing the picture and the clip.
The output is as long as the source clip. Its sound is the source clip’s sound; the model does not generate speech and has no sound-off switch, so mute it in the montage if you do not want it. The proportions of the output follow the picture, not the clip. Connect one picture of the hero, not a sheet of several views.
The cost is shown on the node before the run and is counted by the length of the source clip.
| Setting | Value |
|---|---|
| Model ID | dreamactor |
| Name | DreamActor 2 |
| Availability | available |
| Default for new nodes | no |
Common declared limits
Section titled “Common declared limits”| Setting | Value |
|---|---|
| Capabilities | none |
| Prompt required | no |
| Maximum prompt length | no fixed maximum declared |
| Duration | same as source video; fallback same as source video |
| Audio | origin; fallback origin |
| Aspect ratio | 9:16, 3:4, 1:1, 16:9; fallback 9:16 |
| Declared language support | prompt: English; speech: none |
| Content constraints | people: restriction level: ok |
| Negative prompt strategy | instruction in prompt text |
Mode: motion_transfer
Section titled “Mode: motion_transfer”| Setting | Value |
|---|---|
| Aspect ratio | not declared |
| Combined reference limits | images: 1; videos: 1; audio inputs: 0; combined maximum: 2; audio without visual input: no; video reference seconds: minimum: 1; maximum: 30; video frame: min side: 200 |
| First frame | required |
| Source video | required |
Unavailable modes: avatar_audio, edit_video, recast, text_to_video.
Limits apply together. initial is the studio choice for a new node; default is the fallback for an existing node without that value. Declared language or input support does not guarantee a successful result.
Prompting playbook
Section titled “Prompting playbook”- Give a picture whose framing matches the clip: waist-up picture for a waist-up clip, wide picture for a wide clip. The picture sets the scale of the heroes, and a mismatch makes them look larger or smaller than the performer.
- Use one continuous shot with readable faces and as many performers as there are heroes in the picture. Fast cuts in the clip were not tested.
- Expect no help from words. If the hero must change, change the picture; if the acting must change, change the clip.
Known failure modes
Section titled “Known failure modes”The picture’s framing is kept and the clip’s is not. In the 22 October 2026 trial the picture showed two heroes full-length on a stage while the clip showed the performers to the knee, and the heroes came out in the picture’s scale. The result is therefore not a faithful copy of the clip’s shot, and it is not a way to keep a clip’s background.
The output size is the vendor’s own scaling of the picture rather than the picture’s exact size. Waiting takes a few minutes for a 5–10 second clip, so the node shows progress.
Evidence and limits
Section titled “Evidence and limits”On 22 October 2026 one trial turned two rappers from a ten-second clip into cartoon heroes from a picture. Body movement, expression and lip movement were carried over, the background and framing came from the picture, and the owner rated the result as weak but acceptable. Two further runs of five and ten seconds confirmed that the output length equals the clip length and that the clip’s sound arrives in the output. These are three observations of one scene type: they do not cover real faces, clips with cuts, clips under two seconds, or other pictures.
The vendor’s input schema was read on the same date. It lists an image of up to 4.7 MB and between 480×480 and 1920×1080, a clip of 200×200 to 2048×1440 and up to 30 seconds, and a switch that removes the vendor’s opening second; Vini always removes it. The vendor’s public page does not state a minimum clip length or a release date. See the model–technique applicability matrix for the connected operations and Choosing a model for how availability differs from a rendered outcome.
Reviewed 22 October 2026 against the current studio contract and dated studio research. Vendor documentation checked on that date. Recheck this page when the vendor changes the address, limits or output rules.