Learn a method. Choose a route.
Understand how a technique works, use its real controls, and see what a recipe needs.
Create an outward-and-return video from two generated endpoint images, with independent motion controls and outward-only audio routing.
Read guideCreate a try-on preview from a person photo and one garment reference, configure the actual styling and seed inputs, and understand the provider path and preservation limits.
Read guideCreate a magazine-in-packaging image through two generative stages, with independent cover and presentation settings.
Read guideGenerate one silent video from source frames and a scene prompt, then choose edge guidance, an optional appearance reference, paired sampling settings and playback options.
Read guideMake one silent clip from text and distinguish generation, playback and encoding controls in this exact recipe.
Read guideGenerate images from text, choose between the coupled standard and Lightning configurations, and adjust the controls that reach sampling in this exact recipe.
Read guidePrepare an image request with the retained Nano Banana 2 interface, choose among its actual settings, and understand the limits of its documented preview delivery.
Read guideChoose BiRefNet for a model-backed local alpha workflow or Bria for authenticated hosted background removal, then use the selected recipe’s image preview.
Read guideCreate a new image from text, follow the exact conditioning and sampling path, and adjust real controls without confusing generation with image editing.
Read guideTurn one starting image into a Vidu Q3 video. Choose compatible model and resolution settings, align prompt timing with duration, and control sound and seed with the retained limitations in view.
Read guideStart with source_image in the selected BiRefNet recipe and get a transparent cutout at Sweet Tea Preview Image (output:18). Understand model setup, foreground segmentation and alpha polarity before adapting the result.
Read guideStart with a scene description and a 2048×1024 panorama canvas, choose the standard or Lightning branch, and adjust the controls that actually reach the sampler.
Read guideExtract a separate foreground mask, inspect its alignment and polarity, and decide how an existing edit or composite flow should consume it. Segmentation alone does not preserve original pixels.
Read guideConfigure a silent video from endpoint images and text, understand the shared sampling schedule, and distinguish generation controls from playback and encoding controls.
Read guideStart with an audio-bearing video, generate video-guided music, and receive a mixed video plus a separate music file.
Read guideEnlarge an illustration while controlling initial refinement, tile redraw, seam correction and final sharpening in this exact recipe.
Read guideTurn a written script into one talking-head video; choose the speaker, delivery and video shape separately from export settings.
Read guideStart with one image from text, follow the exact conditioning, sampling and decoding path, and configure size, starting noise and guidance with the recipe's actual controls.
Read guideGenerate an image batch from text and configure this recipe's prompt refinement, style, dimensions and sampling, with graph and interface discrepancies made explicit.
Read guideFor an out-of-memory failure at VAEDecodeTiled node 109 in the selected LTX-2 recipe, try the published smaller_decode_tiles adaptation and inspect the resulting video for seams. This changes spatial VAE decoding only; no measured memory reduction or whole-machine fit is established.
Read guide