FLUX 3 Text to Video: What It Can Do and What Is Still Early Access

Text-to-video is usually the first thing people try when a new model drops. You type a scene and hope for a clip. FLUX 3 from Black Forest Labs includes that path, and the clip can ship with native audio in the same pass.

Note: FLUX 3 Video is live on TheFluxTrain. Clips are 5–20 seconds, 720p or 1080p, with optional native audio. Cheaper 720p drafts can be enhanced to 1080p.

Quick answer: FLUX 3 text-to-video generates a clip up to 20 seconds with native audio from a text prompt. On TheFluxTrain you run it from Video from Text or the FLUX 3 model page. Image generation, BFL's own public API list, and open-weight FLUX 3 Dev are still on the roadmap.

For the full release map, read FLUX 3: Image, Video, Editing Features and Release Status.

What is FLUX 3 text-to-video?

FLUX 3 is a multimodal model. Text-to-video is one input mode: describe a scene, get motion plus sound together.

A lot of older stacks generate silent video and add audio later. BFL's pitch is that dialogue and impacts can stay timed because audio is part of the same generation.

Text-to-video is also the baseline people use for comparisons. BFL's early preference numbers used 10-second, 720p text-to-video clips with audio. Treat those as maker-reported signals until independent tests land.

What can you ask for in a FLUX 3 text-to-video prompt?

From the announced capabilities, a useful prompt usually covers:

• Subject and action: who or what moves, and what they do
• Camera: static, push-in, pan, or cut language if you plan multi-shot later
• Setting and light: place, time of day, atmosphere
• Audio intent: dialogue language, ambient sound, or event sounds
• Duration feel: keep it inside the 20-second one-pass limit

Example

A barista pours latte art in a small Tokyo cafe at golden hour. Camera slow push-in over the counter. Soft jazz in the background. She says in Japanese that the drink is ready. 8 seconds. No cutaways.

You will refine this once you see how the model weights speech versus ambience.

How does FLUX 3 text-to-video fit the wider model?

Text-to-video sits next to:

• Image-to-video from a still or reference
• Video-to-video from a source clip
• Native audio and multilingual dialogue
• 20-second length limits and multi-shot chaining

If you need a locked look before motion, start with a still once FLUX 3 image generation opens, then animate it.

What do you need before you try FLUX 3 text-to-video?

• A TheFluxTrain account and enough credits (Rate Card)
• A clear scene brief, not only mood words
• A duration between 5 and 20 seconds, or Auto
• Reference notes if you will chain more clips later

Time: Prompt writing is minutes. Generation time depends on the model and queue.

What are the limits right now?

• One pass tops out at 20 seconds (5 second minimum on TheFluxTrain)
• Longer stories need multiple clips, not one endless take
• Preference scores vs other models are preliminary and BFL-reported
• FLUX 3 Image stills are not on TheFluxTrain yet; animate any still you already trust

How do you run this on TheFluxTrain?

1. Open Video from Text and pick FLUX 3 Text (or FLUX 3 Text Draft).
2. Set aspect, duration, and 720p or 1080p (drafts stay 720p).
3. Prompt, generate, then optionally extend or enhance a draft from the gallery.

You can also start from the FLUX 3 model page.

Frequently asked questions

Does FLUX 3 support text-to-video?

Yes. BFL lists text-to-video among FLUX 3 Video capabilities, with native audio in the same generation.

How long can a FLUX 3 text-to-video clip be?

Up to 20 seconds in one pass. Longer sequences need chained clips. See the video length guide.

Does text-to-video include sound?

BFL says native audio is part of FLUX 3 Video output, including multilingual dialogue and event-tied sounds.

Is FLUX 3 text-to-video free?

No. On TheFluxTrain it uses credits by duration and resolution. See the Rate Card.

Is FLUX 3 text-to-video on TheFluxTrain yet?

Yes. Use Video from Text or the FLUX 3 model page.

How does it compare to Runway or Luma?

BFL published early preference margins on short text-to-video clips. Independent reviews will matter more once access expands. Start with the release overview for context.

Can I use image references in text-to-video?

FLUX 3 also supports image-to-video and visual references. If your job starts from a still, use the image-to-video guide.

Where is the official announcement?

Black Forest Labs' FLUX 3 post.