FLUX 3 AI Video Generator with Sound

FLUX 3 is Black Forest Labs’ model family for image, video, and audio generation. FLUX 3 Video creates moving scenes with generated sound; FLUX 3 Image generates and edits still pictures. For video, you can start with a text prompt or an image and describe the action that follows. Choose the version that matches the type of content you want to make.

FLUX 3 video generation uses Lumens. Model access and the amount required depend on your account and generation settings. Check the displayed cost before you generate, particularly when changing the clip length or output resolution. You may need a subscription or additional Lumens. PicLumen’s pricing page lists the current plans and allowances.

BFL’s FLUX 3 documentation lists 5–20-second clips for text-to-video and image-to-video generation, at 24 frames per second. The documented workflow renders at HD and uses video upsampling for Full HD, 2K, and 4K outputs. These higher-resolution options should be described as upsampled output. Common framing options include landscape 16:9, vertical 9:16, and square 1:1; check your selected model for the available settings.

Yes. The FLUX 3 workflow documented by Black Forest Labs supports one starting image, a start-and-end pair, or up to ten keyframes. These images define moments within the clip. For a product transition, you might specify a closed package at the beginning and an open package at the end. FLUX 3 generates the movement between those moments. A keyframe sequence is different from supplying general style references.

Black Forest Labs lists English, Spanish, French, German, Japanese, Chinese, Portuguese, and other languages among FLUX 3’s dialogue capabilities. Write the spoken line in the intended language and identify the speaker. A short sentence gives the scene room for pauses and expressions. Listen to the complete result to check pronunciation and speaker assignment, and review lip timing when the person’s mouth is visible.

In BFL's workflow, Draft Mode produces a quicker preview at a lower generation cost. You can choose a draft and use the dedicated enhancement step to render that at full quality. Submitting the prompt as a fresh generation can give you a different result. Draft Mode is useful for comparing creative directions before paying for a full-quality version.

FLUX 3’s continuation mode generates footage after an existing clip, using the source video and audio to guide what happens next. You could use it to carry an entrance into the next action. Changes to footage already recorded use a separate editing workflow. Black Forest Labs documents FLUX Video Edit as a distinct tool for tasks such as changing elements in a source clip.

Start by identifying what went wrong. If the camera circles a product that should rotate in place, specify a fixed camera and a rotating subject. If the clip rushes through several events, reduce the number of actions or allow more time. For image-based generation, check that the source shows the parts needed for the movement. Change one instruction per attempt so you can judge its effect.

FLUX 3 Max is a version of FLUX 3 Video post-trained by fal Research in collaboration with Black Forest Labs. The additional work focuses on visual quality, following prompts, and generation speed. It belongs to the same family, with its own available settings. Performance figures published for Max describe that version and test setup; they should not be applied to every FLUX 3 workflow.