Black Forest Labs Releases Flux 3 Video Model, Claims It Outperforms Seedance 2.0
Black Forest Labs has made its new Flux 3 video generation model generally available, claiming it beats rival Seedance 2.0 in key benchmarks. The model, now accessible via API, generates high-quality video clips from text prompts and image inputs. The company says Flux 3 delivers superior motion coherence, temporal consistency, and visual fidelity compared to competitors.
What Flux 3 Offers
Flux 3 is a text-to-video and image-to-video model. It can generate clips up to 10 seconds long at 24 frames per second. The model supports multiple aspect ratios, including 16:9, 9:16, and 1:1.
Key features include:
- Multi-modal input support. Users can prompt with text alone or combine text with a starting image. The model then animates the image or generates a scene from scratch.
- Advanced motion control. Flux 3 offers better tracking of objects across frames, reducing flicker and unnatural jumps. The company highlights improved handling of fast movements and complex scene transitions.
- High-resolution output. Clips can be rendered at up to 1080p resolution. The model also supports 4x upscaling for select use cases.
Black Forest Labs positions Flux 3 as a direct upgrade to its previous video model, Flux Pro. The new version incorporates architectural improvements and training on a larger, higher-quality dataset.
Performance Claims Against Seedance 2.0
The company published benchmark results comparing Flux 3 to Seedance 2.0, a video model by Chinese startup Seedance. According to Black Forest Labs, Flux 3 beats Seedance 2.0 in three main areas:
“Flux 3 achieves a 12% higher score in visual quality, 15% better temporal consistency, and 18% improved prompt adherence compared to Seedance 2.0,” the company stated in its release.
These figures come from internal evaluations using standard metrics such as FID (Fréchet Inception Distance) and CLIP score. Independent third-party validation has not yet been published.
Availability and Pricing
Flux 3 is now available through Black Forest Labs’ API platform. Developers can integrate it into their own applications. The company offers a pay-as-you-go pricing model:
- Standard tier: $0.02 per second of generated video for 720p output.
- Premium tier: $0.04 per second for 1080p output with higher priority queue access.
A free tier is also available with limited daily usage for testing. The company plans to release an open-source version of Flux 3 in the coming months, following its previous practice with Flux models.
Background: The Video AI Race
Black Forest Labs entered the video generation space in late 2024 with Flux Pro. Since then, the market has become crowded. Competitors include:
- Runway Gen-3 — offers similar length and resolution but is proprietary only.
- Pika Labs — focuses on fast generation and simple UI for consumers.
- Seedance 2.0 — gained attention for its handling of human motion and lighting effects.
Flux 3’s claim to beat Seedance 2.0 aims to position it as the current leader in technical quality. However, many users still report that no video model handles long scenes or complex physics perfectly.
Caveats and Limitations
Black Forest Labs acknowledges that Flux 3, like all current video models, has limitations:
“Flux 3 still struggles with consistent character appearance across very long clips, and fine details like hands or fingers may occasionally be distorted,” the company noted.
The model also requires significant GPU memory. Running even a single generation at 1080p demands around 24 GB of VRAM. That puts it out of reach for many individual creators without cloud API access.
What This Means for Creators
For video producers and AI artists, Flux 3 offers a new tool with real strengths in motion coherence and high resolution. The claimed edge over Seedance 2.0 could make it a go-to choice for projects requiring smooth, cinematic motion.
Yet the landscape changes fast. Seedance may release a counter-update. Open-source alternatives like AnimateDiff continue to improve. The real test will come from user feedback and independent benchmarks over the next few weeks.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.