Black Forest Labs
Flux 3One model for image, video and audio
Flux 3 is Black Forest Labs' new multimodal foundation model, generating image, video, and audio from a single unified architecture. Creations that are truer to life in every style.
FLUX 3 is available on faktry: text-to-video, image-to-video, first/last-frame-to-video, and video extension with native audio, plus text-to-image and image editing up to 4K.
Flux 3 in action
See what Flux 3 can create, generated on faktry.
What is FLUX 3?
FLUX 3 is a multimodal AI foundation model from Black Forest Labs, released on July 23, 2026. It generates image, video, and synchronized native audio from a single unified architecture built on Self-Flow, Black Forest Labs' method for aligning multimodal generation and understanding inside one model. FLUX 3 Video produces clips up to 20 seconds long in a single generation, with multilingual dialogue and sound-visual synchronization.
FLUX 3 at a glance
Last updated: · Model details published by Black Forest Labs
- Developer
- Black Forest Labs
- Released
- July 23, 2026
- Architecture
- Self-Flow multimodal foundation model
- Modalities
- Image, video, and native audio from one model
- Max video length
- 20 seconds per generation
- Audio
- Native, synchronized, multilingual dialogue
- Variants
- FLUX 3 Video, FLUX 3 Image, FLUX 3 Dev, FLUX 3 Action, FLUX-mimic
- Open weights
- Announced but not yet released. FLUX 3 Dev is confirmed as the open-weight variant, with no release date, license or Hugging Face repository published yet
- License
- Not published for FLUX 3. Earlier FLUX.1 [dev] weights ship under the BFL Non-Commercial License
- On faktry
- Live now: FLUX 3 video (text-to-video, image-to-video, first/last-frame-to-video, video extension) and FLUX 3 image (text-to-image, editing up to 4K)
FLUX 3 vs Flux 2
What changes between Black Forest Labs' image-only Flux 2 and the multimodal FLUX 3.
| Capability | Flux 2 | FLUX 3 |
|---|---|---|
| Modalities | Image only | Image, video, and native audio |
| Video generation | Not supported | Up to 20 seconds per generation |
| Audio | Not supported | Native, synchronized, multilingual dialogue |
| Architecture | Image generation model | Self-Flow multimodal foundation model |
| Variants | Pro, Max, Flex | Video, Image, Dev, Action, FLUX-mimic |
| On faktry | Available now | Available now (video and image) |
FLUX 3, Live on faktry
Video billed per second of output, image billed per generation - both live on faktry now.
Video generation
Text to Video
Generate a video with synchronized native audio directly from a text prompt.
Up to 20s · 720p or 1080p · auto or fixed aspect ratio
10 credits / second
Image to Video
Animate a single starting image into a full video clip guided by your prompt.
Up to 20s · 720p or 1080p · native audio
10 credits / second
First/Last Frame to Video
Pin a start frame and an end frame, and let FLUX 3 generate the motion in between.
Up to 20s · 720p or 1080p · native audio
10 credits / second
Extend Video
Continue an existing clip from its final frames, keeping motion and style consistent.
Up to 20s of new footage · 720p or 1080p · native audio
20 credits / second
Duration, resolution, aspect ratio, native audio, and safety tolerance are all configurable per generation.
Image generation
Text to Image
Generate an image directly from a text prompt, with bounding-box composition control over every element.
512sq-4K · 15+ aspect ratios · auto-layout from a single prompt
10 credits (up to 1k) · 150 at 4K
Edit Image
Edit any part of an image while preserving the rest, composing in up to 10 reference images at once.
Up to 10 reference images · box-by-box editing · resolution up to 4K
10 credits (up to 1k) · 150 at 4K
Price scales with resolution: 512sq-1k all cost the same, 2K is 2x, and 4K is 15x - BFL's own cost curve, which is also why 4K renders take longer.
What makes Flux 3 different
Built on Self-Flow technology, one model that understands how the world looks, moves, and sounds.
- Truly Multimodal
- A single model generates image, video, and audio, instead of separate systems bolted together.
- Self-Flow Architecture
- Flow matching aligns generation and understanding across data types, learning how objects hold together, how things move, and how events sound.
- Video with Native Audio
- Generates video with synchronized native audio up to 20 seconds, including text-to-video, image-to-video, and video-to-video.
- Sharper Image Synthesis
- Improved handling of complex prompts and text rendering compared to earlier FLUX models, across styles, aspect ratios, and resolutions.
- Multilingual by Design
- High-accuracy multilingual text rendering in images and multilingual dialogue in generated video.
- Lifelike Motion & Sound
- Strong performance on facial expressions and sound-visual synchronization, with keyframe-to-video control.
Capabilities
A closer look at what Flux 3 can generate, across video and image.
Video Generation
- Text-to-video, image-to-video, and video-to-video
- Native audio, videos up to 20 seconds
- Keyframe-to-video for precise motion control
- Multilingual dialogue with synchronized lip movement
- Diverse visual styles, from photoreal to stylized
- Strong facial expressions and sound-visual sync
Image Synthesis
- Bounding-box composition control - place and resize every element on a 0-1000 coordinate grid
- Pixel-perfect box editing - change one element at a time while everything else stays untouched
- Multi-reference composition from up to 10 images in a single generation
- Improved handling of complex, multi-part prompts
- High-accuracy multilingual text rendering, including hand-lettered signage
- Native 2K and 4K rendering, preserving texture, face, and color detail at full resolution
- Better consistency across generations than earlier FLUX versions
Early Benchmarks
In Black Forest Labs' preliminary text-to-video evaluations, FLUX 3 Video was preferred over other leading video models in head-to-head human comparisons.
77%
vs. Runway Gen-4.5
93%
vs. Luma Ray 3.2
up to 69%
vs. Grok Imagine Video
60%
vs. Kling v3 Pro
Preliminary results published by Black Forest Labs at announcement. Not independently verified by faktry.
Where Flux 3 fits in
One model covering the workflows that used to need several separate tools.
- Content Creation
- Go from a single prompt to on-brand image and video content, in one consistent style.
- Marketing Video
- Produce short video with native audio and multilingual dialogue for campaigns and social.
- Product Visualization
- Turn product concepts into photorealistic images and motion without a separate render pipeline.
- Global Content
- Accurate multilingual text in images and multilingual dialogue in video for international audiences.
Frequently Asked Questions
Still have questions?Flux 3 is Black Forest Labs' new multimodal foundation model. It generates image, video with native audio, and text from one unified model built on Self-Flow technology.
Can I try Flux 3 for free?
Yes. Every new account gets 50 credits, no card required, and both FLUX 3 video and image generation are unlocked from the start.
- 50 credits included, no card required
- Every FLUX 3 video and image operation
- Native audio, up to 20 seconds
- Commercial usage rights