[15:34 Thu,17.July 2025 by blip] |
Adobe announces a comprehensive update to its Firefly AI video model. Among other things, it is said to offer improved motion fidelity in the generated videos, and sound effects can now also be generated for the videos. In addition, thanks to new functions such as composition reference, style presets or cropping tools, users receive significantly more creative control when creating the AI videos. ;) Generate Sound Effects (Beta) Adobe had already given a sneak preview of AI-generated sounds (Project Super Sonic) at the MAX conference in October 2024. Now the option of having suitable sound effects generated directly in the Firefly web app for the calculated video clips is being added to the app as a beta tool. The desired sounds can simply be described via prompt or even demonstrated: short voice recordings can be recorded or uploaded, and Firefly then orients itself on their timing and intensity. The generated SoundFX snippets, as it seems, are each available individually and can be shortened or combined with one another on separate tracks.
Other third-party video models The image and video upscalers from Topaz and Marey from Moonvalley will soon be introduced into Firefly Boards for the supported video models, while Luma AI Ray 2 and Pika 2.2 (which are already included in Boards) will soon also be available in Generate Video. A practical special feature of Adobe Firefly is the ability to use AI models from other providers when generating, above all the best-in-class Google Veo 3. This is doubly clever, because users do not have to purchase lots of different AI subscriptions, but can always choose the one that best solves a task under one interface, and at the same time Adobe can always offer high-quality video results without having to mix at the top with its own model. Even the now updated model seems to score points primarily in the area of landscapes, animals and animations despite improved motion resolution. Realistic human representations, as Veo 3 now does well, are not among its strengths. As long as you have access to other video AIs via Firefly, this is easier to bear, even if it is only with Adobe&s own model that it is ensured that the generated output is commercially usable due to the strictly controlled training material without the risk of violating any rights. Extended video control New tools in video generation are intended to offer more precise control over the calculated images. For example, a video clip can now also be uploaded to Adobe as a composition template. When interpreting the prompt, Firefly then takes over the image composition from the reference video. ;) Typical looks for a video can also be specified via style presets, such as black/white, claymation, anime, lineart or 2D. When uploading desired start and end frames, these can now be freely cropped directly in Firefly. ;) ;) Text to Avatar (Beta) Another new mode in video generation is Text-to-Avatar. This allows clips to be created in which a text script is presented by a figure (an avatar). To do this, you select an avatar template, adjust the background with a color, an image or a video and also select an accent with which to speak. Enhance Prompt Last but not least, Adobe is making the "Enhance Prompt" function available. It is intended to optimize user prompts through clearer formulations so that the Firefly AI can better implement the instructions.  more infos at bei blog.adobe.comdeutsche Version dieser Seite: Neu in Adobe Firefly - Sound Effects, Kompositionsvorlagen und mehr |
 |
|
|