kling-2-6/pro/image-to-video
Turns static visuals into cinematic motion with synced audio and natural camera flow
FLUX 3 Max is a later training pass on Black Forest Labs' FLUX 3 video foundation. The extra work targets sharper pictures, motion that holds together, and closer obedience to the prompt, with a shorter wait than the base model. That is the published aim of FLUX 3 Max.
A preliminary image-to-video board on Artificial Analysis Video Arena reported Elo 1,139 across 4,868 votes. The published note says those confidence intervals overlap other models at the top of that board, so treat the result as a shared top tier. The original FLUX 3 score on the same board was 1,089, a 50-point gap.
A published timing check for a 10-second clip lists 10.95 seconds for this pass, 277.80 seconds for Seedance 2.5 at the same reported Elo, and 126.44 seconds for FLUX 3. Those figures are the source of the "about 22 times" and "about 11.5 times" comparisons. The first published FLUX 3 Max release date is October 19, 2026, and the same note said further training was still underway.
| Advantage | What you get |
|---|---|
| Closer prompt follow | FLUX 3 Max was trained to hold subject, camera, and sound cues more tightly than the base FLUX 3 pass. |
| Shorter published wait | The 10-second check finished in 10.95 seconds, against 126.44 seconds for FLUX 3. |
| Three ways in | FLUX 3 Max can write a scene, start from one image plus an optional last frame, or pin up to ten keyframes. |
| Sound in the same file | Audio is on by default. Turn it off when you want a silent picture. |
Text-to-video on FLUX 3 Max starts from the prompt alone, and auto framing uses 16:9. Image-to-video starts from a still, and auto framing follows that still. Keyframes-to-video places 1 to 10 PNG, JPEG, or WebP images on a 24 fps timeline. Each frame index must be unique, and it cannot pass duration times 24.
On RunComfy this page is the image start. The text-only page and the multi-keyframe page call the same FLUX 3 Max weights with different inputs.
Teams reuse one FLUX 3 Max setup when the same job starts as a script, a still, or a storyboard.
These controls belong to the image-to-video pass of FLUX 3 Max. The text-to-video page drops both image fields. The keyframes page replaces them with up to ten frame-indexed images.
| Parameter | Required | Type | Default | Range / Options | Description |
|---|---|---|---|---|---|
image_url* | Yes (*) | string | Example still | PNG, JPEG, WebP URL | Start frame for FLUX 3 Max on this page. |
prompt* | Yes (*) | string | Example line | Text | How the clip should move after the start frame. |
end_image_url | No | string | Empty | PNG, JPEG, WebP URL | Optional last frame for FLUX 3 Max. |
duration | No | integer | 5 | 5-20 | Length in seconds. |
resolution | No | string | 720p | 720p, 1080p | Output resolution on FLUX 3 Max. 720p is the default. |
reasoning_level | No | string | high | low, high | high spends longer on the prompt. |
aspect_ratio | No | string | auto | auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16 | auto follows the start image here. On a text-only FLUX 3 Max request, auto uses 16:9. |
generate_audio | No | boolean | true | true, false | Create an audio track with the video. |
seed | No | integer | Empty | Integer or empty | Empty means a random seed on FLUX 3 Max. |
Billing follows the length of the generated clip and the resolution you pick. The same FLUX 3 Max rates apply on the text, image, and keyframe pages.
| Resolution | Rate |
|---|---|
| 720p | $0.05 per second |
| 1080p | $0.08 per second |
| Duration | 720p | 1080p |
|---|---|---|
| 5 s | $0.25 | $0.40 |
| 10 s | $0.50 | $0.80 |
| 20 s | $1.00 | $1.60 |
There is no separate per-image fee for a start frame, an end frame, or keyframe stills. A 5-second 720p file is the shortest standard charge on FLUX 3 Max.
Based on publicly reported preliminary evaluations. Length limits below are the published input ranges, not a scored ranking.
| Model | Inputs | Length | Standout |
|---|---|---|---|
| FLUX 3 Max | Text, one image plus an optional end frame, or up to 10 keyframes | 5-20 seconds | Image-to-video Elo 1,139 on the cited preliminary board; published 10-second timing of 10.95 seconds |
| FLUX 3 | Text, image, keyframe, and extend passes on separate pages | Up to 20 seconds; some passes also offer auto length | Earlier foundation. Reported image-to-video Elo 1,089 and a 126.44 second timing on the same 10-second check |
| Seedance 2.5 | Multimodal video generation | Depends on that model's own settings | Same reported Elo in the cited comparison, with a 277.80 second timing for a 10-second clip |
Stay with FLUX 3 when you need an extend-video pass or an auto duration. Use FLUX 3 Max when you want the later training pass and the shorter published wait. This release does not offer extend or auto length.
Turns static visuals into cinematic motion with synced audio and natural camera flow
Transform still images and voice tracks into lifelike talking avatars with precise motion control.
Animate a start image into a cinematic clip with native audio.
Create fluid, expressive animations with multi-shot storytelling features.
MiniMax H3 Max: text or image to video with native audio at 480p/768p/1080p
Text-driven video transformation keeping motion and style consistent across edits.
FLUX 3 Max covers three passes: text-to-video from a prompt, image-to-video from a start image with an optional end frame, and keyframes-to-video from up to ten timed stills. This page is the image-to-video pass.
FLUX 3 Max is a later training pass on the FLUX 3 foundation, aimed at visual fidelity, motion coherence, and prompt adherence. A preliminary image-to-video board reported Elo 1,139 from 4,868 votes, versus 1,089 for FLUX 3. A published 10-second timing was 10.95 seconds, versus 126.44 seconds for FLUX 3.
FLUX 3 Max uses a whole-second length from 5 to 20, with 5 as the default. Resolution is 720p or 1080p. There is no auto-length control and no extend-video pass on this release.
On this image-to-video page, FLUX 3 Max requires a start image and can take an optional end image. For more than two timed stills, use the keyframes pass, which accepts up to ten images.
Yes. Audio is on by default for FLUX 3 Max, so the file can include sound with the picture. Turn Generate Audio off when you want silence. Name the sounds in the prompt if you want them specific.
FLUX 3 Max offers low and high. high is the default and spends more time interpreting the prompt, so generation is slower. low is the faster pass when fewer details have to land.
For FLUX 3 Max, name one camera move, write the action in order, add a small environmental motion, and list a few sounds. On this page the start image already shows the subject, so spend the prompt on what moves next.
Prototype FLUX 3 Max in the RunComfy Playground Web UI, then call the same model through the RunComfy API with the same image, prompt, duration, resolution, and other fields. Jobs spend usd or credits. New accounts usually start with a trial balance.
FLUX 3 Max costs $0.05 per second at 720p and $0.08 per second at 1080p, based on the generated clip length. The same rates apply to text-to-video, image-to-video, and keyframes-to-video. A 5-second 720p clip is $0.25. A 10-second 1080p clip is $0.80.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.
