FLUX 3 Video: Black Forest Labs Claims Top Elo Score at $0.06 Per Second
Agentic AIBlack Forest Labs launched FLUX 3 Video on August 5 — and the capability set it ships with is not what most builders expected from a draft-tier video model at $0.06 per second.
What Happened
FLUX 3 Video is now generally available through Black Forest Labs’ API and select partners, generating HD and Full HD video clips up to 20 seconds. The model supports text-to-video, image-to-video, keyframe generation, and video continuation. Native audio synthesis — including dialogue, sound effects, and ambient noise — is included across all tiers at no additional cost.
The benchmark headline is a 1,135 Elo score for text-to-video and 1,051 for image-to-video, placing FLUX 3 Video first in BFL’s internal testing ahead of Seedance 2.0, Gemini Omni Flash, and Minimax H3. Internal benchmarks warrant the standard caveat, but BFL has typically been conservative in its public claims since the original FLUX image generation launch, and the pricing structure suggests genuine confidence in quality at scale.
Pricing by tier: Draft HD at $0.06/sec for text or image-to-video ($0.12/sec for video-to-video); Full HD at $0.29/sec ($0.53/sec video-to-video). Lip-synced dialogue runs in 14+ languages. The model can also draw on world knowledge for documentary-style content generation.
Why It Matters
The competitive dynamic in AI video generation has compressed dramatically in 2026. Seedance 2.0, Gemini’s video capabilities, and the Chinese open-weight models (Minimax H3, Wan) have all shipped within months of each other. What FLUX 3 Video adds to that field is a native audio layer and a pricing floor that makes production-level generation economically accessible to developers building at scale — not just researchers or enterprise pilots.
The $0.06/sec entry price for HD video with audio changes what’s feasible in automated content workflows. A 20-second draft clip costs $1.20. That’s below the break-even threshold for human video production at almost any quality tier, which means the marginal cost of generating a video variant for A/B testing, localization, or content repurposing approaches zero. Stack Archive has previously tracked how agentic AI production infrastructure changes cost economics — the pricing structure here fits the same pattern.
The 14-language lip sync capability is less headline-grabbing but potentially more strategically significant for media companies running multilingual content operations. Native sync removes a post-production layer that currently requires specialized tools or manual correction.
What This Means for Builders
Developers building video generation pipelines now have a clearly-priced, API-accessible option with documented benchmark performance rather than preview access or waitlist-gated capability. The practical test is whether BFL’s Elo claims hold in production usage patterns — particularly for complex multi-scene prompts — but the price and availability make it immediately testable.
For teams that have been waiting on video generation to become a reliable pipeline input rather than an experimental feature, FLUX 3 Video’s GA status and audio inclusion make it worth evaluating now rather than at the next model release cycle.
The So What
The AI video generation market is no longer about which model exists — it’s about which model can run reliably in a production workflow at a price that justifies the integration cost. FLUX 3 Video’s $0.06/sec floor and native audio put it in that tier. Run it against your specific use case before committing to a Seedance or Gemini integration; the benchmark numbers suggest the gap is closer than the market share numbers imply.
Source: The Decoder
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →