black-forest-labs/flux-3

OfficialView on Replicate →

Generate video with synchronized audio from text, images, or video. FLUX 3 is Black Forest Labs' multimodal model (early access preview).

Capabilities

No capability data available

Cost

Community model (estimated from hardware time)

Input Parameters

Name	Type	Description	Default	Constraints
`prompt`*	string	Text description of the video to generate. Plain language works — the prompt is interpreted and expanded before generation. Describe the scene, action, camera moves, and any audio you want.	`—`	—
`aspect_ratio`	string	Aspect ratio of the generated video. 'auto' picks a ratio from your prompt and any inputs.	`"auto"`	auto21:92:116:94:31:13:49:16
`draft`	boolean	Generate a fast, low-cost draft preview instead of a full-quality clip. Drafts are 720p.	`false`	—
`duration`	string	Length of the generated clip in seconds. 'auto' lets the model pick to fit the content. Three or more images need an explicit duration.	`"auto"`	auto567891011121314151617181920
`generate_audio`	boolean	Generate synchronized audio (ambient sound, speech, effects). Set to false for a silent clip.	`true`	—
`images`	array	Optional images to drive the video, placed on screen pixel for pixel. One image opens the clip (image-to-video). Two images start and end it. Three or more (up to 10) become a storyboard — the first starts it, the last ends it, and the rest fall evenly in between (needs a duration). Leave empty for text-to-video. Must be PNG, JPEG, or WebP.		—
`resolution`	string	Output resolution.	`"720p"`	720p1080p
`safety_tolerance`	integer	Moderation tolerance for input and output, 0 (strictest) to 4 (most permissive). Requests with image or video inputs are limited to 2.	`2`	min: 0, max: 4
`start_video`	string(uri)	A video to continue from its final frames. Use this to extend a shot or chain generations into a longer sequence. Must be an mp4, at most 50MB and 15 seconds. Can't be combined with images.	`—`	—

promptrequiredstring

Text description of the video to generate. Plain language works — the prompt is interpreted and expanded before generation. Describe the scene, action, camera moves, and any audio you want.

aspect_ratiostring

Aspect ratio of the generated video. 'auto' picks a ratio from your prompt and any inputs.

Default: "auto"

auto21:92:116:94:31:13:49:16

draftboolean

Generate a fast, low-cost draft preview instead of a full-quality clip. Drafts are 720p.

Default: false

durationstring

Length of the generated clip in seconds. 'auto' lets the model pick to fit the content. Three or more images need an explicit duration.

Default: "auto"

auto567891011121314151617181920

generate_audioboolean

Generate synchronized audio (ambient sound, speech, effects). Set to false for a silent clip.

Default: true

imagesarray

Optional images to drive the video, placed on screen pixel for pixel. One image opens the clip (image-to-video). Two images start and end it. Three or more (up to 10) become a storyboard — the first starts it, the last ends it, and the rest fall evenly in between (needs a duration). Leave empty for text-to-video. Must be PNG, JPEG, or WebP.

Default:

resolutionstring

Output resolution.

Default: "720p"

720p1080p

safety_toleranceinteger

Moderation tolerance for input and output, 0 (strictest) to 4 (most permissive). Requests with image or video inputs are limited to 2.

Default: 2min: 0, max: 4

start_videostring

A video to continue from its final frames. Use this to extend a shot or chain generations into a longer sequence. Must be an mp4, at most 50MB and 15 seconds. Can't be combined with images.

Version: 93515f53aad7Updated: 9/20/202627.8K runs