Black Forest Labs halves 2K and 4K video prices

Black Forest Labs has cut native 2K and 4K generation prices for FLUX 3 Video by 50% and says rendering is now up to twice as fast. The company attributes both changes to backend compute optimizations, according to its announcement.

The lower rates reduce the cost of high-resolution iteration, particularly for long takes and continuation jobs. Black Forest Labs also positions FLUX 3 Video against Kling 3 on quality, price, and latency, although independent benchmark coverage remains limited.

The new 4K math

FLUX 3 Video is Black Forest Labs’ first multimodal video foundation model. Its Self-Flow architecture supports text-to-video, image-to-video, keyframes, and video continuation within one system. It generates clips up to 20 seconds long at resolutions up to 4K, with synchronized speech, sound effects, and ambience produced in the same pass.

Native rendering means the model generates at the requested resolution directly instead of relying on a separate upscaler. Black Forest Labs added native 2K and 4K output to its API on September 10, 2026, initially charging $0.40 per second for QHD and $0.80 per second for UHD.

A 20-second 4K generation therefore falls from $16 to $8. A 20-second 4K video-to-video or continuation job, previously priced at $0.95 per second, falls from $19 to $9.50 under the same reduction.

BFL targets Kling on price and quality

Kling 3 has established a strong premium offering with native 4K output at 60 frames per second and clips up to 15 seconds long. Artificial Analysis lists Kling’s 720p Standard tier at about $0.25 per second and its 1080p Pro tier at roughly $0.34 per second. Third-party providers typically charge a premium for 4K, so exact comparisons depend on the provider, mode, and included features.

Black Forest Labs reports that evaluators preferred FLUX 3 over Grok Imagine Video in up to 69% of comparisons, Kling v3 Pro in 60%, and Happy Horse v1 in 59%. FLUX 3 recorded a 52% preference rate against both Seedance 2.0 and Gemini Omni Flash. These vendor-run evaluations provide an early quality signal, but differences in prompts, sampling settings, clip selection, and evaluator pools can materially affect preference scores.

Faster renders change the workflow

Serving optimizations can cut latency by up to half without requiring a new API endpoint. Combined with lower per-second pricing, that change makes higher resolutions more practical during development and review.

• 2K and 4K can enter the iteration loop. Teams can review framing, texture, and fine motion at delivery resolution earlier instead of reserving high-resolution generation for the final render.
• Draft Mode lowers exploration costs. Developers can generate an HD preview for $0.06 per second, approve its composition, and use Draft Enhance to rebuild the video at full quality.
• Audio remains part of each generation. Native multilingual speech, effects, and ambience can remove a separate audio-generation and synchronization step.
• Existing composition controls remain available. Supported aspect ratios include 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16, with first-frame and last-frame steering.
• Current integrations keep the same endpoint. Developers can test the revised tiers without routing requests to a separate API, though model identifiers and account limits still warrant verification.

Test the whole pipeline

FLUX 3 Video remains API-only, rounds billable duration up to the next whole second, and has less independent benchmark coverage than its vendor-published preference results. Model weights are unavailable, so teams that require self-hosting, custom inference infrastructure, or offline deployment will need another option.

A production evaluation should measure more than headline price and generation time:

• Prompt adherence and motion consistency across representative workloads
• Visual drift during 20-second takes and continuation jobs
• Speech quality, lip synchronization, effects, and ambience
• Median and tail latency under expected concurrency
• Retry rates, failed-job billing, and whole-second rounding
• Total cost across drafts, enhancements, continuations, and final renders

Teams using Kling or Runway for premium video now have a stronger cost case for an A/B test, especially when native audio and 20-second takes reduce downstream work. Black Forest Labs lists the current rates on its pricing page and publishes model capabilities on the model page.