- Home
- What Is FLUX 3
GuidesUpdated
What Is FLUX 3? FLUX 3 Video Model and Its Family
Black Forest Labs' FLUX 3: video with sound, plus Max, Image, Klein and Action.
FLUX 3 is Black Forest Labs' multimodal model, announced on July 23, 2026: one model trained on images, video and audio. It is best known for video with synchronized sound, with dialogue, effects and ambience in clips up to 20 seconds, and it also has image and robotics versions. Is FLUX 3 open source? Only two parts, FLUX 3 Image Klein and FLUX 3 Action, have open weights so far. Below: release dates, the five models, and where to try them.
FLUX 3 at a glance
What is FLUX 3, in eight lines. Checked against Black Forest Labs' own pages and Hugging Face on November 6, 2026.
- Maker
- Black Forest Labs, the team behind FLUX.1 and FLUX.2
- Announced
- July 23, 2026
- What it makes
- Video with sound, images, and robot actions
- One backbone learns from images, video and audio together; each FLUX 3 model is built on it.
- Video length
- Up to 20 seconds a clip, at 24 fps
- Resolution
- HD up to 4K on Black Forest Labs' API
- 720p or 1080p on this site, with 2K and 4K through an upscale pass.
- Sound
- Dialogue with lip sync, effects and ambience, made with the picture
- Open weights
- FLUX 3 Image Klein and FLUX 3 Action only
- Where to use it
- Black Forest Labs' API and dashboard, apps that license it, or a browser tool like this one
The FLUX 3 family: Video, Max, Image, Klein, Action and Dev
The name FLUX 3 covers one shared backbone and the FLUX 3 models built on it, and most of the confusion online comes from mixing them up. Each row says what that model makes, when it arrived, whether you can download it, and whether it runs here. You will also see FLUX 3 Turbo on this site: it is this site's own fast preview tier, not a Black Forest Labs model.
| Model | What it makes | Released | Weights | On this site | |
|---|---|---|---|---|---|
| FLUX 3 VideoThe video model most people mean | Video with sound from text, images or an existing clip | July 23, 2026 | Hosted only | Yes, as FLUX 3, plus FLUX 3 Video Edit | Try it |
| FLUX 3 MaxFaster, closer prompt following | Video with sound, 5 to 20 seconds, tuned for speed | October 19, 2026 | Hosted only | Yes | Try Max |
| FLUX 3 ImageThe full image model | Images and edits up to 4K, with layout boxes and up to 10 references | October 2026 | Hosted only; commercial weights through BFL sales | No | |
| FLUX 3 Image KleinSmall open-weight image editor | Images and photo edits from a 5.6B-parameter model | October 13–15, 2026 | Open, on Hugging Face | Yes, as FLUX 3 Klein | Try Klein |
| FLUX 3 ActionRobot control | Robot actions from a 7B-parameter world action model | September 2026 | Open, on Hugging Face | No | |
| FLUX 3 DevThe open backbone, announced | Downloadable weights of the shared FLUX 3 backbone | Coming soon, no date | Not out yet | No |
What the FLUX 3 video model can do
FLUX 3 Video is the FLUX 3 video model most people mean. These are the abilities Black Forest Labs lists for it; the comparison further down marks which ones this site offers.
Sound in the same pass
Dialogue in several languages with lip sync, sound effects and ambience come out with the picture, so a clip needs no separate voice or foley step. You write the spoken words into the prompt.
Several shots from one prompt
One prompt can describe a short sequence of shots with cuts between them, which suits a small scene or an ad with a setup and a payoff.
Up to 20 seconds a clip
A single FLUX 3 clip runs up to 20 seconds at 24 fps, long enough for a spoken line, a reaction and a closing beat.
Start, end and key frames
Start the clip on a photo, pin the last frame too, or guide it with up to 10 keyframes when the motion has to hit set poses.
Edit or extend a clip
Change one thing in footage you already have, such as the weather, a prop or a line of dialogue, or continue a clip past its last frame.
Text that reads
Black Forest Labs says words inside the frame, such as signs, labels and titles, render accurately, which matters for product and brand shots.
Is FLUX 3 open source?
Partly. As of November 6, 2026, two FLUX 3 models have open weights on Hugging Face: FLUX 3 Image Klein, a 5.6B image-editing model, and FLUX 3 Action, a 7B robot-control model. FLUX 3 Video, FLUX 3 Max and the full FLUX 3 Image are hosted only. Black Forest Labs says open weights for the FLUX 3 Dev backbone are coming soon, with no date.
Which FLUX 3 open weights are on Hugging Face
Black Forest Labs' Hugging Face page lists FLUX 3 Image Klein in a distilled and a base (undistilled) version, plus a ComfyUI build, and FLUX 3 Action in base, so101 and droid versions. Klein runs on a single consumer GPU, which makes it the FLUX 3 model you can run locally today. There is no repository for FLUX 3 Video, FLUX 3 Max or the full FLUX 3 Image.
When FLUX 3 Dev open weights are due
Black Forest Labs' FLUX 3 FAQ says open weights for FLUX 3 Dev are coming soon and gives no FLUX 3 Dev release date. Until they land, a FLUX 3 Dev download, a GGUF file or a local FLUX 3 video workflow cannot exist yet; for local video you wait, and for local images you use Klein.
What the FLUX Kommunity License allows
Klein and Action share the FLUX Kommunity License v1.0. Non-commercial use is free. If you and your affiliates earn under 5 million US dollars a year, you count as a qualifying user and may use the outputs commercially; larger companies need a commercial licence from Black Forest Labs. So FLUX 3 open weights means you can download and run these models under those terms, not that FLUX 3 is open source in the usual software sense.
FLUX 3 release dates
Each FLUX 3 release date so far and whether you can use that part now, from Black Forest Labs' blog and model pages and the Hugging Face repositories.
- July 23, 2026
- FLUX 3 launch, video with sound first
- Usable now through Black Forest Labs' API and dashboard, and on this site.
- September 2026
- FLUX 3 Action, with open weights
- Weights on Hugging Face since September 22.
- October 13–15, 2026
- FLUX 3 Image Klein, with open weights
- On Hugging Face from October 13; Black Forest Labs announced it on October 15.
- October 2026
- FLUX 3 Image, the full image model
- Hosted on Black Forest Labs' API and Playground.
- October 19, 2026
- FLUX 3 Max
- Hosted only; runs on this site.
- No date yet
- FLUX 3 Dev open weights
- Black Forest Labs lists them as coming soon.
Is FLUX 3 good? Where it ranks in November 2026
On the Artificial Analysis video leaderboard of November 5, 2026, FLUX 3 Max (1156, #4) and Seedance 2.5 (1155, #5) sit in one overlapping group for text to video, just behind MiniMax H3 (1160, #3), with FLUX 3 at 1139 (#8). For image to video, Seedance 2.5 leads at 1159, about 90 points ahead of FLUX 3 (1069) and FLUX 3 Max (1067). The scores are Elo ratings built from user votes and they move as votes come in. In short, FLUX 3 is at its strongest when you start from words; when you start from a photo, Seedance 2.5 currently rates higher.
| FLUX 3 MaxTry FLUX 3 Max | FLUX 3FLUX 3 vs other models | Seedance 2.5FLUX 3 vs Seedance 2.5 | |
|---|---|---|---|
| Text to video score | 1156 (±16) | 1139 | 1155 (±14) |
| Text to video standing | #4, overlapping with the top group | #8, inside the same range | #5, overlapping with the top group |
| Image to video score | 1067 | 1069 | 1159 (±15) |
| Image to video standing | About 90 points behind | About 90 points behind | #1 |
FLUX 3 from Black Forest Labs vs FLUX 3 on this site
This site is an independent browser tool, not affiliated with Black Forest Labs, and it runs part of the FLUX 3 family. Where a feature is missing here, the table names the closest thing that works.
| Black Forest Labs API and dashboardDeveloper account, pay per second | This siteIn the browser, free sign-up, creditsMake a FLUX 3 video | |
|---|---|---|
| Text or image to video | Yes | Yes |
| Start and end frame | Yes | Yes |
| Up to 10 keyframes | Yes | Partly: FLUX 3 takes up to 10 images in order, evenly spaced |
| Extend a clip | Yes | No |
| Edit a clip | Yes | Yes, with FLUX 3 Video Edit, clips up to 15 seconds |
| Resolution | HD up to 4K | 720p or 1080p; 2K or 4K with FLUX Video Upscale |
| Sound | Yes | Yes, on FLUX 3 and FLUX 3 Max |
| Cheap preview first | Draft mode | FLUX 3 Draft Mode, on this site's FLUX 3 Turbo preview tier (silent) |
| FLUX 3 Image, full model | API and Playground | No |
| FLUX 3 Image Klein | Open weights | Yes, as FLUX 3 Klein |
| FLUX 3 Action | Open weights | No |
Try FLUX 3 here
Start with the how-to guide if this is your first FLUX 3 clip, or try FLUX 3 straight from a model page.


How to Use FLUX 3 Video: Access It and Make Your First Clip
How to use FLUX 3: where to use FLUX 3 and how to access it, then make your first FLUX 3 video step by step, with what each try costs and what is free.


FLUX 3 Prompt Guide: Write Video Prompts with Real Examples
A FLUX 3 prompt guide for video: write a FLUX 3 video prompt with camera, sound and dialogue, and copy FLUX 3 prompts shown with the clips they made.


FLUX 3 Image to Video: Turn a Photo Into a Clip, Free to Try
Turn a photo into video with FLUX 3 image to video: upload a still, describe the motion, get a 2–20 second clip. Sign up free; 10 credits make your first clip.


FLUX 3 Max: Turn Text or a Photo into Video with Sound
Run FLUX 3 Max text to video or image to video with dialogue and ambient sound, 5–20 s at 720p or 1080p. See same-prompt tests timed on this site.


FLUX 3 Draft Mode: Fast Preview, Then Full Quality
Use FLUX 3 Draft Mode to preview a text-to-video or photo-to-video idea on FLUX 3 Turbo, then re-render the same shot at 720p or 1080p on FLUX 3.
FLUX 3 questions
FLUX.2, from November 2025, is Black Forest Labs' image generation and editing family. FLUX 3 is one model trained on images, video and audio together, so it adds video with synchronized sound and a robotics branch, FLUX 3 Action. On the image side, FLUX 3 Image Klein takes over from the small FLUX.2 Klein models as BFL's open-weight image model.
Yes. Black Forest Labs announced FLUX 3 on July 23, 2026, and today you can use FLUX 3 Video and FLUX 3 Image through its API and dashboard, or FLUX 3 and FLUX 3 Max on this site. The part still to come is the FLUX 3 Dev open weights, which BFL lists as coming soon.
Only part of it. FLUX 3 Image Klein runs on a single consumer GPU from its Hugging Face weights, including a ComfyUI build, and FLUX 3 Action can be downloaded for robotics work. FLUX 3 Video and FLUX 3 Max are hosted only, so running FLUX 3 video locally waits on the FLUX 3 Dev weights.
Yes. FLUX 3 Video generates dialogue, sound effects and ambience with the picture, and you put the exact spoken words in your prompt. On this site, FLUX 3 and FLUX 3 Max make sound, on by default; the FLUX 3 Turbo preview tier is silent.
FLUX 3 Turbo is this site's own fast preview tier, not a Black Forest Labs model. It makes silent clips of 2 to 15 seconds at up to 1080P, so you can check framing and motion cheaply before a full render with sound on FLUX 3 or FLUX 3 Max.
Not fully, anywhere. Black Forest Labs charges per second of video. On this site new accounts get 10 free credits — one 2-second clip on FLUX 3 Turbo at 480P, without sound, or five 1K FLUX 3 Image Klein stills; signed out, you can try 3 Turbo Image stills a day — and FLUX 3 or FLUX 3 Max with sound starts at 175 credits. The open-weight Klein image model is free to run on your own GPU for non-commercial use.
No. This site is an independent product and is not affiliated with Black Forest Labs. It gives you FLUX 3, FLUX 3 Max, FLUX 3 Video Edit, FLUX Video Upscale and FLUX 3 Klein in the browser on one credit balance, alongside its own FLUX 3 Turbo preview tier.
Make a FLUX 3 clip now
Describe one shot, preview it on FLUX 3 Turbo, then render it with sound on FLUX 3 or FLUX 3 Max. First time? The How to Use FLUX 3 guide walks through your first clip step by step.
