Current product
Generate through a named live model
The Studio shows the selected model, settings, and exact credit estimate before creating the task.
Open StudioCurrent AI video Studio · FLUX 3 tracked separately
Direct a publishable AI video from one brief
Generate now with Wan 2.5, Seedance 2, or Kling v3. FLUX 3 remains a clearly labelled Early Access watchlist until a real endpoint is connected and verified.
Failed generations auto-refundedFLUX 3 Video is not available in this studio yet. Current generations use the model selected in Studio.
Current model and cost baseline
Derived from the same registry used for generation
Wan 2.5Fast
5/10s · 720p/1080p · three aspect ratios
50
credits · 5s · 720p
Seedance 2Pro
5/10s · 720p/1080p · three aspect ratios
125
credits · 5s · 720p
Kling v3Cinematic
5/10s · 720p/1080p · three aspect ratios
200
credits · 5s · 720p
FLUX 3 Video
Black Forest Labs marks FLUX 3 as Early Access. It is not connected to AI Flux3, so this row cannot create a task or consume credits.
Product facts reviewed 2026-08-03
Generate a clip
Direct the shot from one workspace
Choose a live model, set the frame and duration, then see the exact credit estimate before generation. FLUX 3 stays clearly marked until API access opens.
Quick prompts
Native audio
Include dialogue, ambience, and effects when supported.
Estimated cost: 50 creditsAutomatic refund if generation fails
See plans and save on more generationsFailed generations auto-refunded
Credits never expire
Commercial use subject to provider terms
Evidence before adjectives
What is live, what is official, and what is still unknown
AI Flux3 does not label an empty frame or third-party demo as its own output. Use the current Studio now, and follow the cited BFL sources for FLUX 3 Early Access.
Recoverable work
Find every task in My Videos
Pending, completed, failed, and refunded task context stays attached to the signed-in owner.
View My VideosOfficial boundary
Track FLUX 3 without pretending it is live
The status page cites Black Forest Labs and separates announced capabilities from what AI Flux3 can run today.
Read statusNo fabricated users, testimonials, countdowns, or generated-output claims.Read the official BFL announcement
Built around the brief
One studio, four publishing workflows
01
Filmmaker
Previsualize camera moves, bridge keyframes, and test a sequence before committing a crew.
02
Marketer
Turn a campaign idea into platform-ready 16:9, 9:16, and square variations.
03
Educator
Make short visual explanations with spoken context, labels, and clear pacing.
04
Indie Creator
Produce consistent clips without juggling separate prompt, audio, and export tools.
Three decisions
From a written beat to a directed clip
01
Write the beat
Name the subject, action, camera move, environment, and sound. Quick prompts show the expected level of detail.
02
Set the frame
Choose the model, aspect ratio, resolution, duration, and audio before credits are calculated.
03
Generate and review
Submit one clear brief. Failed generations are refunded automatically, so the balance reflects usable output.
Published capability baseline
Official FLUX 3 scope vs this Studio today
Black Forest Labs describes FLUX 3 in Early Access. The right column is the narrower capability currently exposed and billed by AI Flux3.
| Capability | FLUX 3 Video | Available in AI Flux3 |
|---|---|---|
| Maximum clip length | Up to 20 seconds | 5 or 10 seconds |
| Audio generation | Native synchronized audio | Selected models only |
| Keyframe transitions | Designed for transitions | No keyframe control |
| Multilingual lip-sync | Multilingual dialogue | No dedicated lip-sync control |
| Typography in frame | Improved text rendering | No verified typography mode |
Model architecture
A shared representation for image, video, and sound
FLUX 3 is described by Black Forest Labs as a unified multimodal model trained jointly across images, video, and audio. Instead of treating motion and sound as separate finishing steps, the representation learns visual appearance, temporal change, and acoustic events together. Text, image, and video inputs can therefore condition the same generation process, while the model maintains relationships between a subject, its movement, the camera, dialogue, and surrounding sound over time. That joint training is the technical basis for longer clips, native audio, keyframe transitions, multilingual lip synchronization, in-frame typography, and multiple connected clips. The same representation is also designed to extend toward action prediction, giving the model a more coherent account of what changes from one moment to the next.
This is a product-level summary of BFL’s public FLUX 3 announcement, not a claim that AI Flux3 operates the FLUX 3 model today.
Credits
Video credits sized for real production
Start with 1,000 credits, move up when your publishing cadence does. Annual plans include two months at no extra cost.
Lite
$9.90/mo
For a weekly social clip
- 1,000 credits per month
- 720p and all aspect ratios
- Native audio when supported
- Failed generations auto-refunded
- Credits never expire
- Commercial use subject to provider terms
Pro
$29.90/mo
For a steady publishing calendar
- 3,500 credits per month
- 720p and 1080p output
- Native audio when supported
- All available frontier models
- Failed generations auto-refunded
- Credits never expire
- Commercial use subject to provider terms
Max
$59.90/mo
For campaigns and longer sequences
- 7,500 credits per month
- 720p and 1080p output
- Native audio when supported
- All available frontier models
- Failed generations auto-refunded
- Credits never expire
- Commercial use subject to provider terms
FAQ
Questions before your first render
Clear answers about the models available now, credit handling, commercial use, and the FLUX 3 early-access timeline.
01What is AI Flux3?
AI Flux3 is an independent AI video workspace. It lets you prepare text-to-video, image-to-video, and continuation briefs, then choose an available model, frame, duration, and audio setting from one Studio.
02Can I generate videos here today?
Yes. Signed-in users with enough credits can generate through Wan 2.5, Seedance 2, and Kling v3. The selected model and exact credit estimate are shown before submission.
03Are you affiliated with Black Forest Labs?
No. AI Flux3 is an independent project and is not affiliated with, endorsed by, or operated by Black Forest Labs. FLUX is a trademark of Black Forest Labs Inc.
04Which video models are listed in Studio?
Studio lists Wan 2.5 as the fast cost tier, Seedance 2 as the primary production tier, and Kling v3 as the cinematic tier. FLUX 3 appears only as a disabled Early Access option.
05Which generation modes are supported?
The Studio supports text-to-video, first-frame image-to-video, and video continuation. Images use the authenticated upload flow; continuation currently accepts a public video URL.
06Does every model generate native audio?
No. Audio capability varies by model and endpoint. The audio switch requests dialogue, ambience, and effects only when the selected provider supports them.
07What clip lengths and formats can I choose?
The current Studio supports 5- and 10-second requests at 720p or 1080p, with 16:9 YouTube, 9:16 TikTok, and 1:1 square frames.
08How are generation credits calculated?
Credits are calculated from the selected model’s per-second rate, duration, resolution, and audio setting. The Generate button shows the estimate before a request is submitted.
09What happens when a generation fails?
The system claims credits when it creates a generation task and restores the same amount when the provider rejects the task or the task reaches the failure timeout. The task remains visible in My Videos.
10Can I use generated videos commercially?
A commercial-use license is included for output generated through paid plans, subject to the selected model provider’s terms and your responsibility for prompts, source media, trademarks, and depicted people.
11How is my prompt and upload data handled?
Prompts and uploaded inputs are used to fulfill the generation request and operate your account. Do not submit confidential, personal, or rights-restricted material; the Privacy Policy explains the current handling boundary.
12When can I use FLUX 3 itself?
FLUX 3 Video is currently in API early access for initial partners. It is not available in AI Flux3 today. The disabled Early Access entry marks the planned integration point; availability will change only after a production API opens to us.