Text-to-video generation with Marey allows for a completely net new video to be generated, rather than beginning from a keyframe. We recommend reading through our prompting best practices to help structure your prompts.
How it works
By default, Marey controls on Fal will be displayed as a form but can be switched to an editable JSON, or a read-only text snippet in your preferred langauge. To change the page layout, click 'Form' (default) in layout dropdown menu.
Here's a breakdown of the inputs you'll need to set for text-to-video. For demo purposes, we'll stick to the default 'Form' layout:
Prompt: This is where you'll write your prompt in the Prompt description box. Be descriptive about camera details (e.g. 35mm, handheld shot), scene or environment, subjects and action of your video. Prompts with at least 50 words provide best results.
Additional Settings: There are a number of additional settings to customize your generations including dimension size, duration, and seed value.
Dimension: Select your preferred video dimension from the drop down menu. By default, videos will be 1920x1080.
Duration: Select either five or 10 second video output. By default, videos will be five seconds.
Negative Prompt: Add or remove negative prompts to help guide the model away from undesirable results.
Seed: Select your desired seed value. Seed determines the randomness and different possibilities of each generation. By default, seed will be set to '9'. Use '-1' for a random seed value.
Preview and Export
Once you click generate your video will begin being processed. Once the generation is finished, the video will be displayed in the 'Result' panel with a 'Download' and 'Share' button below the video
You can also switch the results to JSON, which will provide you with a URL that can be copy and pasted into a browser for easy sharing and download.
