How to extend an AI video in Seedance through the Syntx aggregator

What does it mean to extend a video in Seedance?

PCreating a video in Seedance 2.0 is a function that allows you to add new frames to a ready-made video and logically continue the scene. Instead of creating a video again, you take the final fragment of the original video, describe what should happen next, and the Seedance neural network generates a continuation while preserving the main details:

• incharacters' tenderness;

• withimage style;

• aatmosphere of the stage;

• dcamera movements;

• ogeneral visual direction.

TThis approach is especially useful when you need to extend a video with a neural network, but save the already created material and not start generation again. To work, you can use an aggregator Syntx (Syntx), where the Seedance neural network is available along with other video creation tools. This allows you to run generation in one interface without having to switch between different services.

>>> PAYBACK - promotional code for a 15% discount when replenishing your balance 

Step 1: Prepare a video segment for extension

PThe first thing you need to do before renewing your video online through Seedance is to prepare the last fragment of the original video. It is necessary to cut the final part of the video, which will be used as the basis for the sequel.

NFor example, our video about how the Wolf Shaman runs along the glacier and jumps into the abyss ends with this shot: 

NWe want to show where he jumped and what’s inside this abyss. Therefore, we cut out the last piece of the video, which begins from the moment it is pushed off the ground, shows the moment of take-off and the beginning of the fall:

There are several important requirements:

• the duration of the fragment must be at least 4 seconds - this is a limitation of the Seedance model;

• all the elements that need to be preserved in the new continuation should be clearly visible in the frame;

• Characters, objects and details of the scene should be displayed as clearly as possible.

For example, if the video contains a person, a car, and a specific location, all of these elements should be included in the final fragment. If some object is difficult to see or is missing in the selected segment, it is better to add it separately as a reference.

It is important to take into account the limitations of Seedance: you cannot use real videos with real people as reference. However, restrictions do not apply to anthropomorphic animals and cartoon characters.

Step 2. Prepare additional references

HTo get a more accurate continuation, you can use additional reference images. This is not a mandatory stage: the Seedance neural network can only work with a text description of the scene, but additional materials help to better control the result.

For example, you can prepare:

• kamouth of a new location;

• andimage of the future environment;

• nan example of an object that should appear in the frame;

• via visual guide for the character's next action.

Dlet's put it down, after the extension, our Wolf should end up on the seabed, surrounded by jellyfish and luminescent corals. To create a location map, go to the “Text” tab and select Claude. Throw it in  instructions for generating element maps. 

CFollowing the instructions, we describe in free form what kind of map of the area we want to receive. 

KAs you can see, after our request, a ready-made industrial product immediately appears, which we copy and send to  tab “Design” and with the help of Nano Banana we get a map of the future area in which our Wolf will find himself.

These references allow for a more manageable result, especially when maintaining a specific composition or style is important.

Step 3. Create a prompt to extend the video

CThe next stage is preparing a description of what should happen next.

HTo extend a video using AI, it is not enough to write a short phrase like “continue the scene.” The more accurately the continuation is described, the better the neural network understands the task. To create a prompt, a ready-made template for video extension is used. 

The template must be inserted into the chat with Claude, and then describe the future scene. The description should include:

• towhat action does the character continue;

• toHow does the camera change?

• Thuo happens to the environment;

• towhat objects appear;

• towhat mood should be maintained.

If additional images are used, they must be specified inside the prompt using special tags:

• @image1;

• @image2.

TIt should be placed next to the first mention of the element to which the reference refers.

For example:

"The character enters a new location @image1 and continues to move forward...” For us it looks like this (the @image1 tag refers to the image of the general plan of the underwater world): 

PIf necessary, in the same prompt Claude can be asked to change the finished prompt:

• withshorten or increase timing;

• dadd more details;

• atclarify the use of specific references;

• andchange the sequence of actions.

TThis approach helps to prepare a prompt that is suitable specifically for generating videos through Seedance.

Step 4. Set up Seedance in the Syntx aggregator

PAfter preparing the video and prompt, we move on to generation. Open the “Video” section in the Syntx aggregator and select the Seedance neural network. Next, you need to configure the generation parameters.

Anduse the following settings:

• Bmodel version: Seedance 2.0;

• Rmode: Pro - for maximum quality of the result;

• Caspect ratio: selected to match the source video, for example 16:9;

• DLength: Specify the required continuation length.

PAfter that, download:

• nprepared fragment of the original video;

• dAdditional reference images, if used;

• goriginal prompt.
  
  

  

  
   

Step 5. Check the linking of references before generation

PAfter loading images, you must manually attach tags in the prompt text. Initially, text will be displayed instead of the image: @image1, @image2, etc. After correct binding, a small plate with a thumbnail of the image should appear next to it. This means that Seedance understands which reference refers to a specific description element.

Important: It is enough to attach tags only at the beginning of the prompt, where the images used are decrypted. In other parts of the text you can leave the usual notation: @image1

> It is also worth remembering that when copying and pasting a prompt, the attached images may become unpinned. Before each generation, be sure to check the tags and add them again if necessary.

Step 6: Combine the original video and the continuation

PAfter generation is complete, you will receive a new fragment that continues the original video. Where did our Wolf Shaman end up:

dNext is the last stage when you need to connect the two parts:

1. original video;

2. sequel created through Seedance.

DTo do this, you can use any editing program. After combining, you will get a full-fledged video with a smooth continuation of the plot.

Why Seedance is suitable for video extension via Syntx

Seedance 2.0 allows you not just to create individual video fragments, but to develop existing scenes. With the help of proper preparation of the initial frame, additional references and an accurate prompt, you can:

• withkeep the style of the original video;

• ncontinue character movement;

• rexpand the plot;

• andhandle generation interruption without re-creating the entire video.

And thanks to the Syntx aggregator, working with AI for video becomes easier: the necessary neural network is available in one interface, and the process of creating a video that can be extended online is faster and more convenient.

___________________________________________________________

Syntx web version, 15 tokens for free

Telegram bot Syntx

PAYBACK - promotional code for a 15% discount when replenishing your balance