FLUX 3

One multimodal model for ImageVideoAudioAction-Prediction
Creations are truer to life in every kind of style.

Try it

One model,
multiple modalities.

Stylistically diverse beyond just cinematic, with native audio and up to 20 second clips in a single generation. Start from text, an image, or keyframes, and get multiple shots in one take.

cheetah running across the savanna
Sample 1 of 9 for Video

Introducing FLUX 3

FLUX 3 Video’s core capabilities.

Handles simple or complex prompts.

Example 1 of 9 for Text to video

Draft mode.

Explore creative directions fast at a fraction of the cost. A draft generation returns a fast preview of your prompt. When a draft is right, send it back and FLUX 3 will render the same video at full quality.

Koi drifting between floating paper lanterns at night, draft render

Draft

Fast generation at a fraction of the cost.

Koi drifting between floating paper lanterns at night, full-quality render

Normal

Renders the draft in full quality.

Already in production.

What teams are saying about FLUX 3 Video.

  • FLUX 3 puts a new medium of entertainment and storytelling in our users’ hands, and it responds well to an agent’s direction. Hermes can generate a shot, checks it, then continues from it, so those 20-second clips become cohesive longform pieces.”
    Dillon Rolnick, CEO, Nous Research
  • FLUX 3 is a significant addition to our next generation of video tools, including Picsart's flagship AI Playground. Having motion and audio grounded in the same model lets us treat each clip as a self-contained shot and handle narrative and timing at the pipeline level. Our users expect access to the latest and greatest models, and FLUX 3 raises the bar for what they can create with AI.”
    Mikayel Vardanyan, COO & Co-Founder, Picsart
  • With FLUX 3, we have the opportunity to boldly explore new directions and experiment with ideas that were previously unimaginable. The model delivers such high visual quality that we can bring our IPs to life in fully on-brand videos — without the time and cost constraints of traditional productions. This unlocks a whole new space for innovation and opens up entirely new possibilities.”
    Rebecca Gottwald, Chief AI Officer, Burda Media
  • Most video models are built for film. FLUX 3 goes beyond that, into the work our users also spend their days on: product, brand, motion design. That range in a single model is why we wanted it on Magnific from day one.”
    Omar Pera, CPO, Magnific
  • We've been building with FLUX 3 ahead of launch, and it changes what our Creative Pros can do with video. The quality bar has moved to a point where the ideas that used to get cut for budget or time reasons are now on the table. We're seeing our community go from concept to polished, on-brand video in hours instead of weeks. For a platform built around empowering creatives, that's exactly the kind of step change we want to put in their hands.”
    Hichame Assi, CEO, Envato
Testimonial 1 of 5

Use cases

Content
Content
Virtual-try-on
Virtual-try-on
Storyboarding
Storyboarding
cinematic western shot of a saddled horse galloping through the desert
Filmmaking
Advertising
Advertising
whiteboard explainer animation of a hand drawing dot clusters with a marker
Explainer Videos
street-food chef flame-cooking in a wok at a night market
Documentary
World simulation
World simulation
Worldbuilding
Worldbuilding
a wolf running through a snowy forest
Nature & Wildlife
anime-style schoolgirl on a rooftop overlooking a city at sunset
Anime
liquid droplets forming a logo reveal on black
VFX & Titles
a harpist performing inside a glowing blue ice cave
Music & Performance
retro CRT terminal booting with green monospace text
Text and Typography

Each modality makes the others better.

FLUX 3 builds on Self-Flow, our approach for efficiently aligning multimodal generation and understanding within the same underlying architecture.

Two-panel figure comparing Self-Flow with Flow Matching. Self-Flow comes out ahead in both: lower generation error across the modalities on the left, higher manipulation success rate through finetuning on the right.
Self-Flow vs. Flow Matching (FM). Left: generation error (Fréchet distance) per modality, each normalized to FM = 100 (lower is better). Right: success rate on manipulation tasks averaged over four task groups through finetuning (higher is better).

See what people are making.

A growing collection of videos created by the FLUX 3 community.

  • I have been playing around with Flux 3 and it is really good at a wide range of prompts. @NousResearch @bfl_ai

    Ryan Fogleon X

  • Future Vibes AIon X

  • Flux3 videos have this subtle, mystical charm. The results really depend on how you tweak the prompts, and it's honestly pretty fascinating.#flux3 #bfl

    Jin.Bon X

  • Okay - If you are into filmmaking and care at all about video/audio generation, I am telling you now... This partnership between Hermes Agent/Flux3 is incredible. Been testing it out for a few hours and already, it's awesome. I can imagine where this goes. Don't sleep on this.…

    SamJWassermanon X

  • an older man sitting in a dry grass field

    seriously impressed by how flux 3 can do dialogue with super vague prompts like "ranting about ai"

    Dreaming Tulpaon X

  • a dinosaur riding a skateboard in a suburban street

    I think flux 3 is my new favorite model

    kvickarton X

  • a breakdancer on cardboard on a New York City street with taxis

    FLUX 3 for me feels like a better SORA. I think people are going to make some really wacky stuff with this model. “Girl from the attached ref images is break dancing on a flattened cardboard box on a bustling NYC sidewalk in 1993”

    AIWarperon X

  • close-up of hands kneading dough

    FLUX 3 made a 20 second cooking tutorial. Lots of multishot in the footage here! It really loves putting butter everywhere 😂.

    Jerrod Lewon X

  • workers in an ancient stone quarry

    FLUX 3 prompt: "How pyramids are made? Visualize it."

    gokayfemon X

  • a shelf of Black Forest Labs FLUX 3 chocolate bars

    FLUX 3 has been really versatile so far! Heres a first person POV shot at the grocery store.@bfl_ai When are we getting Early Access to the Candy Bars Tho? 🫶😂

    realrebelaion X

  • an armored rider racing a hover-speeder in a packed stadium

    FLUX.3 image-to-video pushed into full action mode: three speeders, collisions, laser fire, consistent direction, and one continuous 20-second tracking shot

    HBCoop_on X

  • a French Revolution street uprising with tricolor flags

    More FLUX 3 time travel gopro footage! This time it's the French Revolution and WW2 D-Day invasion.@bfl_ai

    cfryanton X

  • 1990s sitcom scene with anthropomorphic dogs in a living room

    More @bfl_ai FLUX 3 gens

    ostrisaion X

  • surreal sitcom scene with a wooden mannequin figure in a living room

    Przetestowałem najnowsze narzędzie od @bfl_ai - Flux 3. Powiedziałbym że coś pomiędzy seedance 2 a sora . Przynajmniej z polskim sobie radzi. Więcej przykładów niżej:

    STARPARon X

  • a person building a primitive structure in the jungle

    hellorobon X

  • claymation stop-motion figure drinking coffee in a kitchen

    "claymation story about a man consistently getting smushed by a large giant hand that enters the scene so he cant live his life" Flux<3

    SyntaxDiffusionon X

  • a swarm of legged robots marching across a snowy landscape

    Seriously considering producing this as a full National Geographic documentary. Flux 3 is insanely good at these, this was made with one prompt!

    venturetwinson X

  • a person in a red jacket walking along an autumn sidewalk

    This is absolutely amazing with Flux 3’s split-screen rendering. More power to storytelling! Upscaled with Topaz Astra. Prompt : Split-screen video. Two equal vertical halves. Both halves show the SAME event, at the SAME time, frame-synchronized, filmed by two different…

    umesh_aion X

  • a first-person view riding through city streets in a wooden cart

    New FLUX 3 early access test - treating the model like a time machine! Samurai duel in a burning Kyoto temple during a clan war, first person gopro footage.@bfl_ai

    cfryanton X

  • a figure walking through misty crop fields at dawn near pagodas

    flux 3 prompt: "Daily life of the Aztecs"

    gokayfemon X

  • a giant creature walking down a city street at night among police cars

    One of the most cinematic FLUX 3 videos I've generated. Could genuinely see this in a movie scene.@bfl_ai

    cfryanton X

  • close-up of a person in black leather in a cyberpunk city action scene

    Testing camera-relative continuity. FLUX 3 kept the rider close to the lens while the wet city accelerated around her, maintaining her identity, jacket detail, and blue headlights on the bike glow through the entire shot.

    HBCoop_on X

  • a woman eating an apple shown from four consistent angles

    FLUX 3 split camera testing This is insane

    realrebelaion X

Frequently asked questions

What is FLUX 3?

FLUX 3 is Black Forest Labs' multimodal foundation model for generating and understanding video, audio, images, and actions within one unified architecture.

What can I use as input?

FLUX 3 Video supports text-to-video, image-to-video, video-to-video, video continuation, and controlled transitions using keyframes.

How long and at what resolution can FLUX 3 generate video?

FLUX 3 can generate clips up to 20 seconds long in a single generation, in HD (up to 1 megapixel per frame) or FHD (up to 2 megapixels per frame). Resolution bands are set by total pixels per frame, not aspect ratio. Draft mode is HD only.

Does FLUX 3 generate video and audio together?

Yes. FLUX 3 can generate optional native audio with the video, including multilingual dialogue, synchronized speech, sound effects, and environmental ambience. Audio is included at no extra charge.

How much does FLUX 3 Video cost?

FLUX 3 uses pay-as-you-go pricing with no subscriptions or seat fees. Text or image to video costs $0.06/sec for Draft HD, $0.17/sec for standard HD, or $0.29/sec for standard FHD. Video to video costs $0.12/sec for Draft HD, $0.41/sec for standard HD, or $0.53/sec for standard FHD. Five seconds of standard text-to-video in HD costs $0.85. View pricing

What is FLUX 3 Draft mode?

Draft mode generates a faster, lower-cost HD preview. Once the direction is right, the draft can be enhanced into a full-quality FHD result. Draft Enhance is priced at the corresponding standard FHD rate.

How can I access FLUX 3?

FLUX 3 Video is available through the Black Forest Labs dashboard and API on a pay-as-you-go basis. Enterprise customers can request volume discounts, custom pricing, SLA guarantees, and dedicated support. Talk to sales

Are FLUX 3 Image, Action, and open weights available?

FLUX 3 Image, FLUX 3 Action, and the FLUX 3 Dev open-weight backbone are separate parts of the rollout. FLUX 3 Image generates and edits images, FLUX 3 Action is rolling out through selected research and commercial partners beginning with mimic robotics, and open-weight access to the FLUX 3 Dev multimodal backbone is coming soon. Read the rollout plan

Make things with FLUX 3.

One model built for exploration and production.

Try itRead more