Flux
Image model family for accurate text and photoreal stills, plus FLUX 3 video with native audio
FLUX is Black Forest Labs' family of image models, now joined by FLUX 3, which adds video generation with synced speech and sound. Image models are sold per output image through a pay-as-you-go API, and the FLUX.2 open-weights model can be self-hosted and fine-tuned under a license. An official hosted MCP server lets Claude and other assistants call FLUX directly.
- 14 creators using
- 4 news items tracked
- 2024 first launched
- Freemium pricing
- Moderate learning curve
What Flux can do
- Infinite canvas No
- AI agent No
- Node / flow builder No
- Public API Yes
- MCP server Yes
19 of 26 capabilities confirmed
| Capability | Supported | Details |
|---|---|---|
| Text to image | Yes | FLUX.2 and FLUX 3 Image |
| Image editing | Yes | FLUX.2 editing, FLUX Kontext, FLUX Tools (outpainting, erase, deblur, virtual try-on) |
| Text to video | Yes | FLUX 3 Video, up to 20 seconds with audio |
| Image to video | Yes | FLUX 3 Video can start from an image or keyframes |
| Video to video | Partial | Video upscale (2K/4K) and video edit listed in FLUX Tools; no general restyle mode found |
| Native audio in video | Yes | FLUX 3 Video generates audio alongside video |
| Lip-sync | Yes | Lip-synced speech in multiple languages from FLUX 3 Video |
| Voice / TTS | Partial | Speech is generated inside video output only; no standalone TTS or voice cloning found |
| Music & SFX | Partial | Ambient sound and sound effects inside video output; no music generator found |
| Upscaling | Yes | FLUX Video Upscale to 2K and 4K (Aug 2026) |
| Character consistency | Partial | FLUX.2 multi-reference composition (up to 10 images, per ComfyUI launch post); FLUX 3 reference inputs announced as upcoming |
| Camera control | Partial | Multi-shot and camera-angle switching within one generation; no explicit camera-move parameters listed |
| Custom training | Yes | Fine-tuning and LoRA rights on open-weights licence tiers; fine-tuned FLUX.2 endpoints in API public beta |
| Third-party models | No |
| Capability | Supported | Details |
|---|---|---|
| Infinite canvas | No | No canvas or moodboard workspace found on bfl.ai; the product is API, Playground and open weights. |
| Node / flow builder | No | No native node builder found. ComfyUI runs FLUX.2 as a third-party node workflow. |
| AI agent | No | |
| Chat assistant | No | |
| Team collaboration | Not confirmed |
| Capability | Supported | Details |
|---|---|---|
| Public API | Yes | Pay-as-you-go API with per-model pricing and docs at docs.bfl.ai |
| MCP server | Yes | Official hosted FLUX MCP server at https://mcp.bfl.ai |
| Mobile app | Not confirmed | |
| Desktop / local | Yes | Open-weights FLUX.2 models run locally; Klein runs on-device on ASUS ProArt laptops |
| Open source | Partial | FLUX.2 [dev] and FLUX 3 Action are open weights under licence; several variants are non-commercial only |
| Capability | Supported | Details |
|---|---|---|
| Free tier | Partial | Non-commercial open weights are free to download; no free API credits found on the pricing page |
| Commercial use | Partial | Open weights are non-commercial unless licensed; licensed tiers and Enterprise allow commercial use. API output terms not verified. |
Yes Partial No Not confirmed by the vendor
What makes Flux different
- 01
Open weights: FLUX.2 [dev] and the klein models can be downloaded, self-hosted and fine-tuned, with LoRA rights on the licensed tiers. Midjourney is closed.
- 02
Pay per image with no subscription: FLUX.2 [klein] 4B from $0.014 per image, FLUX.2 [pro] from $0.03. Midjourney and Adobe Firefly sell monthly plans.
- 03
Text rendering and high-resolution output: FLUX 3 Image is priced by size from $0.041 (768x768) to $0.607 (about 16 MP, 4K).
- 04
Official hosted MCP server at https://mcp.bfl.ai, with setup steps for Claude, Claude Code, Cursor, Codex and Windsurf.
- 05
Native audio video: FLUX 3 Video generates speech, sound effects and ambience in the same pass, which Midjourney's image-to-video does not offer.
Latest from Flux
- 2w ago Flux 3 Image Klein Adds Multi-Reference Generation and 5.6B Parameter Open Weights Black Forest Labs released Flux 3 Image Klein, a 5.6B parameter model that improves text-to-image accuracy and multi-reference consistency for creators.
- Aug 20 FLUX Upscale Adds Native 2K and 4K Video Upscaling via Dedicated Endpoint Black Forest Labs launched FLUX Upscale for video, allowing creators to regenerate clips at native 2K and 4K resolutions.
- Aug 4 FLUX 3 Video, Part 1: Generation FLUX 3 Video, Part 1
- May 7 How Our Forward Deployed Engineers Work with Customers Black Forest Labs launches a forward-deployed engineering program to help creators and studios integrate FLUX models into custom production pipelines and infrastructure.
About Flux
FLUX is the image model family from Black Forest Labs. The current lineup is FLUX.2 (pro, flex, max and the small klein models, plus open-weights dev), FLUX 3 Image, FLUX Kontext for editing, and the FLUX Tools for targeted edits such as outpainting, erasing and virtual try-on. FLUX.1 models are now labelled previous generation on the pricing page.
What changed in 2026 is mostly the move beyond stills. FLUX 3 was announced on 23 July 2026 as a multimodal family covering video, image, audio and action prediction. FLUX 3 Video generates clips up to 20 seconds with native audio, including speech and lip-sync in several languages, and it can start from text, an image or keyframes. The FLUX 3 Image model was still described as coming to early access in the weeks after the announcement. FLUX Video Upscale (2K and 4K) followed on 20 August 2026.
For a film workflow, FLUX is most useful for key art, concept frames, environment plates and text-heavy graphics, where its text rendering is a real differentiator. It is priced per image, so it suits projects that generate a few hundred frames rather than thousands. The open-weights FLUX.2 models fit studios that want a local pipeline in ComfyUI and custom LoRAs, though commercial use of the open weights requires a licence tier. Video is newer and its availability and pricing are less settled, so verify before building a shot list around it.
What's new in Flux
-
FLUX 3 Action: open-weights 7B world action model
Black Forest Labs released an open-weights 7B model for robot and game action prediction, which it says ranks first on the RoboLab benchmark.
Source: bfl.ai ↗ -
FLUX Video Upscale: 2K and 4K
A video upscaling option for FLUX video outputs at 2K and 4K, added to the FLUX Tools line.
Source: bfl.ai ↗ -
FLUX 3 Video generally available through the API
Clips up to 20 seconds with native audio, 720p and 1080p (1080p via upscaling). Draft mode is offered at a lower cost. Pricing is not published.
Source: bfl.ai ↗ -
FLUX 3 family announced
Black Forest Labs announced FLUX 3 as one multimodal model family covering video, image, audio and action, with FLUX 3 Video in early access at launch.
Source: bfl.ai ↗ -
FLUX.2 Klein runs on device
ASUS ProArt laptops gained support for the FLUX.2 Klein models for local, on-device generation.
Source: bfl.ai ↗
Models available in Flux
-
FLUX 3 Video Latest $ Black Forest Labs Up to 20 s with native audio, 720p/1080p; pricing not published
-
-
-
$ · $$ · $$$ = relative cost against other models of the same type, on fal.ai list prices. Cost inside Flux depends on its own plans and credits.
Key Features
- FLUX 3 Video: clips up to 20 seconds with native audio, speech and lip-sync in several languages
- FLUX 3 Image: text-to-image with accurate text rendering and output up to about 4K
- FLUX.2 [pro], [flex], [max] API tiers priced per image by output size
- FLUX.2 [klein] 4B and 9B small models for cheap or local generation
- FLUX Kontext [pro] and [max] for instruction-based image editing
- FLUX Tools: outpainting, erase, deblur and virtual try-on
- FLUX Video Upscale to 2K and 4K
- Open-weights FLUX.2 [dev] and klein for self-hosting in ComfyUI
- Fine-tuned FLUX.2 endpoints in API public beta
- Official hosted MCP server at mcp.bfl.ai
When to reach for it — and when to skip
Reach for it when…
- Strong text rendering and photoreal stills, which matter for title cards, signage and key art
- Pay per image with no subscription, so light users pay only for what they generate
- Open weights and LoRA rights mean a studio can build a local pipeline with its own styles
- Native-audio video from FLUX 3 Video is a capability Midjourney does not offer
- Official MCP server and a documented API make it easy to call from agent workflows
Skip it when…
- Video is new. FLUX 3 Video availability and FLUX 3 Image early access have been described differently across sources, and video pricing is not published
- Open-weights models are non-commercial by default, and commercial use needs a licence quoted by sales
- No canvas, node builder or creative agent, so multi-step boards and storyboards must be built elsewhere
- Per-image cost climbs quickly at 4K. FLUX 3 Image is $0.607 per 4K image
- Smaller LoRA and community ecosystem than Stable Diffusion
Best For
✓ Ideal for
- Key art and concept frames where legible text in the image matters
- Environment plates and set-design variations at 2K or 4K, priced per megapixel
- Studios that want open weights and custom LoRAs running locally in ComfyUI
- Previs with native-audio clips of up to 20 seconds, once video access is confirmed for your account
- Pipelines that call image generation from an AI assistant through the MCP server
✗ Not built for
- Creators who want a canvas, storyboard agent or one-click ad builder (use Krea, Higgsfield or Freepik instead)
- High-volume generation on a flat budget, where a subscription is cheaper
- Commercial use of the open weights without a licence
Working Tips from Filmmakers Using Flux
- 01 Start with FLUX.2 [klein] 4B for previs at $0.014 per image, then switch to FLUX.2 [pro] or 3 Image for hero frames. Output size drives the cost, so render at the size you need, not 4K by default
- 02 For text-heavy frames, put the exact words in quotes in the prompt and check the result at full size before using it
- 03 The open weights are free only for non-commercial work. Get a licence quote before a paid project uses FLUX.2 [dev]
- 04 Use FLUX Video Upscale on a 720p draft only when the shot needs 2K or 4K, since upscaling adds cost on top of generation
Flux pricing in 2026
| Plan | Price | What you get |
|---|---|---|
| Open weights, non-commercial | Free one-time download |
|
| API pay-as-you-go | From $0.014 per image pay-per-image |
|
| Builder (open-weights licence) | Contact sales licence |
|
| Platform (open-weights licence) | Contact sales licence |
|
| Professional (open-weights licence) | Contact sales licence |
|
| Enterprise | Contact sales custom |
|
Prices checked against Flux's pricing page on 9 Oct 2026. Vendors change plans often, so confirm before you buy.
Flux vs Midjourney vs Adobe Firefly vs Leonardo.ai
| Flux | Midjourney | Adobe Firefly | Leonardo.ai | |
|---|---|---|---|---|
| Price | Pay per image on the API, from $0.014 (FLUX.2 klein 4B). Open weights are free for non-commercial use; commercial open-weights licences are quoted by sales. | From $10/mo ($8/mo billed annually); 20% off on annual plans | Free daily generations, or Standard from $9.99/mo (2,000 credits); Pro Plus and Premium have a 30% first-year promo ending Oct 21 | See the vendor's pricing page |
| Infinite canvas | No | Partial | Yes | Yes |
| Node / flow builder | No | No | Not confirmed | Partial |
| AI agent | No | No | Yes | Not confirmed |
| Chat assistant | No | No | Yes | Not confirmed |
| Team collaboration | Not confirmed | Partial | Yes | Partial |
| Text to image | Yes | Yes | Yes | Yes |
| Image editing | Yes | Yes | Yes | Yes |
| Text to video | Yes | Not confirmed | Yes | Yes |
| Image to video | Yes | Yes | Yes | Yes |
| Video to video | Partial | No | Yes | Not confirmed |
| Native audio in video | Yes | Not confirmed | Not confirmed | Partial |
| Lip-sync | Yes | No | Not confirmed | Not confirmed |
| Voice / TTS | Partial | No | Yes | Not confirmed |
| Music & SFX | Partial | No | Yes | Not confirmed |
| Upscaling | Yes | Yes | Yes | Partial |
| Character consistency | Partial | Partial | Partial | Partial |
| Camera control | Partial | Not confirmed | Not confirmed | Partial |
| Custom training | Yes | Partial | Yes | Yes |
| Third-party models | No | No | Yes | Yes |
| Public API | Yes | No | Yes | Yes |
| MCP server | Yes | No | Not confirmed | Not confirmed |
| Mobile app | Not confirmed | Not confirmed | Not confirmed | Yes |
| Desktop / local | Yes | No | Partial | No |
| Open source | Partial | No | No | No |
| Free tier | Partial | No | Yes | Yes |
| Commercial use | Partial | Yes | Yes | Yes |
Each column comes from that tool's own researched page. "Not confirmed" means the vendor doesn't say.
Open the full Flux vs Midjourney vs Adobe Firefly vs Leonardo.ai comparison →Use Cases
Integrations
Tags
Creators Using Flux
AI filmmaker converting ideas to story that sells and move
3D ARTIST, AI FILMMAKER
AI Filmmaker • AI Automation Specialist • Cinematic Storyteller • Generative AI Creator
AI FILMAKER
Sources
Facts on this page were checked against Flux's own pricing, product and changelog pages on , then cross-checked with independent coverage where vendor pages were silent. Anything we couldn't confirm is marked "Not confirmed" rather than guessed.
- Pricing page bfl.ai
- API pricing per model (docs) docs.bfl.ai
- Black Forest Labs blog (changelog of 2026 announcements) bfl.ai
- FLUX 3 announcement bfl.ai
- FLUX 3 Video, Part 1: Generation bfl.ai
- Licensing tiers bfl.ai
- What is the FLUX MCP? help.bfl.ai
- FLUX 3 Video pricing guide layer3labs.io
Spotted something out of date? Tell us and we'll re-check it.
Discussion
No comments yet — be the first.