LTX 2 Retake
Neural Network
LTX 2 Retake is a model for precise video editing: changing actions, camera angles, and specific fragments of a finished frame.
Specifications
- Input
- text, images
- Output
- video
Video generation
(duration 8 sec., 720p)
Video generation
(duration 8 sec., 1080p)
How it works LTX 2 Retake?
LTX 2 Retake doesn't work from a blank slate but with an already finished frame: you select a scene fragment and rework just that part—rephrase the description, change the action in the frame, or choose a different camera angle. The rest of the take remains in place, so you don't have to reassemble the entire clip for one bad second. This approach saves time when the general idea is already found and edits concern details. In BotHub, the model is available from Russia without a VPN or foreign card; payment is made in rubles using Russian cards and is based on actual usage—for consumed caps, which do not expire. Nearby, in one window, there are over 250 neural networks: you can assemble a rough cut with one model and refine it with this one. For production, there is a unified OpenAI-compatible API, AES-GCM encryption, and corporate terms with a contract, invoice, and admin panel. Useful for editors fixing specific shots in a finished assembly, marketers creating versions of one ad frame for different audiences, content creators reshooting an angle without losing the found scene, and producers quickly showing a client an alternative action in a frame.Frequently asked questions about LTX 2 Retake
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a Lightricks video model: it creates video from text descriptions or animates an uploaded image. Suitable for ad inserts, dynamic social media scenes, video pre-visualization, and quick idea testing. In BotHub, it is available from Russia without a VPN or foreign card, with payment in rubles.
You get a video clip as output—the model generates a moving image, not text. We do not specify resolution and duration parameters in our data, so refer to the technical header of the page and the results of your first generations: create a test clip and evaluate the image yourself.
A separate step-by-step reasoning mode is not noted in our data, and it is not the main thing for a video model: a precise prompt is more important here. Describe the scene, camera movement, lighting, and pace—the more specific the shot setup, the more predictable the result. It is more convenient to entrust logical chains to text models in the same BotHub window.
There is no mention of function calling or structured output in our data, so do not rely on them for integration. However, the model works via the unified OpenAI-compatible BotHub API: you send video generation requests just like with other models and switch between them without rewriting code.
Images—yes: this is the image-to-video mode, where a frame becomes the starting point for animation. This is convenient for keeping the desired character, background, or frame style. Input of audio and video is not noted in our data, so plan your work around two scenarios: text description and reference image.
Our data marks the model as ENGLISH_ONLY, so it is better to compose prompts according to the rules accepted for it and not rely on free input in any language. The BotHub interface itself is in Russian: you place requests as usual, pay in rubles, and switch to another model if necessary.