FLUX 3 Max is a video generation model from the FLUX family by Black Forest Labs: a post-trained variant of FLUX 3 Video, tuned for stronger prompt adherence and better visual quality and optimized for low latency. It makes 5 to 20 second clips at 720p or 1080p, with optional generated audio, from a text prompt, from a starting image, or from up to 10 keyframe images pinned to moments in the clip.
Below: what the three modes do, every request parameter, what it costs on apimodels.app, how fast it is in our own tests, how it compares with Seedance 2.5, Kling V3, Veo 3.1 and MiniMax H3 Max Turbo, and when another model is the better pick.
FLUX 3 Max is a latency-optimized, post-trained build of FLUX 3 Video by Black Forest Labs. Its publisher reports two things: it ranks joint first in a preliminary Artificial Analysis image-to-video evaluation, statistically alongside the highest-ranked models, and it generated a 10-second video in about 11 seconds in their benchmark, against 277.8 seconds for Seedance 2.5. Both are the publisher's figures; preliminary rankings change as votes come in, and real completion time depends on settings and load.
All three modes use the same model id, flux-3-max, and the same endpoint. What you send decides the mode. A prompt alone is text-to-video. Add first_frame_url and it becomes image-to-video: the image sets the opening composition and the prompt directs what moves; add last_frame_url to guide the final frame as well. Send a keyframes array instead and it becomes keyframes-to-video.
Keyframes-to-video pins 1 to 10 images to positions in the clip. Each keyframe is {image_url, frame_index}, where frame_index is a frame number in the 24 fps output: 0 is the first frame and duration × 24 is the last, so frame 120 is the end of a 5-second clip. On apimodels.app you can also write {image_url, time} in seconds and it is converted to the frame number for you. Positions must be unique, and keyframes cannot be combined with first_frame_url or last_frame_url.
Every parameter except prompt is optional. Unknown or out-of-range values return a 400 with the reason before anything is charged.
| Parameter | Values | Default |
|---|---|---|
| model | flux-3-max | required |
| prompt | Text describing the scene, motion, camera and sound | required |
| first_frame_url | Image URL or base64 (PNG, JPEG, WebP); turns on image-to-video | none |
| last_frame_url | Image for the final frame; needs first_frame_url | none |
| keyframes | 1 to 10 of {image_url, frame_index} or {image_url, time} | none |
| duration | 5 to 20 whole seconds | 5 |
| resolution | 720p, 1080p | 720p |
| aspect_ratio | auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16 | auto (16:9 for text, follows the first image otherwise) |
| generate_audio | true, false | true |
| reasoning_level | low, high (low is faster) | high |
| seed | Integer, for repeatable results | random |
| callback_url | HTTPS URL that receives the result when the task finishes | none |
Billed per second of video, by resolution. The price is the same for all three modes, and keyframes, first and last frames and audio cost nothing extra. Until 00:00 UTC on 7 November 2026 a special price applies: $0.085 per second at 720p and $0.145 per second at 1080p. Only successful generations are charged; failed or rejected requests are refunded automatically.
| Resolution | Per second | 5 s clip | 10 s clip | 20 s clip |
|---|---|---|---|---|
| 720p | $0.221 | $1.105 | $2.21 | $4.42 |
| 1080p | $0.377 | $1.885 | $3.77 | $7.54 |
| Special price until Nov 7: 720p | $0.085 | $0.425 | $0.85 | $1.70 |
| Special price until Nov 7: 1080p | $0.145 | $0.725 | $1.45 | $2.90 |
We ran each mode once through the public API on 20 October 2026, at 720p and 5 seconds with reasoning_level set to low, timing from the create request to a finished, downloadable file. One run per mode is a spot check, not a benchmark; times vary with load and settings.
| Mode | Input | Time to finished file | Output |
|---|---|---|---|
| Text-to-video | Prompt only, 16:9 | 9 s | 1280×704, 24 fps, with audio |
| Image-to-video | Square first frame | 15 s | 960×960, follows the image, with audio |
| Keyframes-to-video | 2 keyframes at 0 s and 5 s, audio off | 19 s | 960×960, no audio track |
All five are on apimodels.app under the same API key, so switching is a change of model id. FLUX 3 Max is the pick when turnaround matters: fast iterations, 1080p, clips up to 20 seconds, and keyframe control at up to 10 points. Seedance 2.5 is the pick for long, reference-heavy work: up to 30 seconds in one pass, up to 50 reference assets, and editing or extending existing video, at 480p or 720p. At 720p, FLUX 3 Max's regular price ($0.221/s) is below Seedance 2.5 ($0.270/s); Seedance 2.5 at 480p ($0.120/s) and Kling V3 standard ($0.12/s) start at a lower price per second.
| FLUX 3 Max | Seedance 2.5 | Kling V3 | Veo 3.1 | MiniMax H3 Max Turbo | |
|---|---|---|---|---|---|
| Price on apimodels.app | $0.221/s (720p), $0.377/s (1080p) | $0.120/s (480p), $0.270/s (720p) | $0.12/s standard, $0.16/s pro (+$0.06/s with audio) | $1.00 per 8 s clip | $0.06/s (480P), $0.096/s (768P) |
| Length | 5–20 s | 4–30 s | 3–15 s | 8 s | 5–15 s |
| Resolution | 720p, 1080p | 480p, 720p | Standard and Pro tiers | 1080p | 480P, 768P |
| Image control | First + last frame, or up to 10 keyframes | Up to 50 reference assets | Image-to-video | Start and end frame | First + last frame |
| Audio | Generated by default, can be turned off | Audio references and audio-track editing | Optional, billed per second | Synced audio | Not supported |
| Strength | Fast turnaround, keyframes | Long clips, references, edit and extend | Per-second pricing, audio | Google video quality | Clips in seconds at low resolution |
apimodels.app accepts Alipay as well as cards and PayPal, so you do not need an overseas credit card or a separate account with the model publisher. Create a key in the console, top up, and call the same endpoint; results are delivered as links you can download directly.
Choose another model if you need clips longer than 20 seconds (Seedance 2.5 goes to 30 seconds in one pass), many reference images to keep a character or product consistent across shots (Seedance 2.5 takes up to 50 reference assets), editing or extending an existing video (Seedance 2.5), 4K output (Gemini Omni Flash), reference video for motion transfer, or a starting price for rough drafts (LTX 2.3 is from $0.02 per second).
cURL
# Text-to-video: create the task
curl -X POST https://api.apimodels.app/v1/video/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "flux-3-max",
"prompt": "A red paper boat drifts down a rain-soaked city gutter at dusk, neon reflections ripple on the water, camera follows at street level, soft rain sounds",
"duration": 5,
"resolution": "720p",
"aspect_ratio": "16:9"
}'
# -> {"code":200,"data":{"taskId":"cm...","state":"pending"}}
# Poll until state is "completed", then download data.resultUrls[0]
curl "https://api.apimodels.app/v1/video/generations?task_id=TASK_ID" \
-H "Authorization: Bearer YOUR_API_KEY"Python
# Image-to-video with a first and last frame, then wait for the result
import time, requests
API = "https://api.apimodels.app/v1/video/generations"
H = {"Authorization": "Bearer YOUR_API_KEY"}
task = requests.post(API, headers=H, json={
"model": "flux-3-max",
"prompt": "The camera slowly moves closer as the mug turns on the table, steam rises, quiet kitchen ambience",
"first_frame_url": "https://example.com/first.jpg",
"last_frame_url": "https://example.com/last.jpg", # optional
"duration": 8,
"resolution": "1080p",
"reasoning_level": "low", # faster
}).json()["data"]["taskId"]
while True:
d = requests.get(API, headers=H, params={"task_id": task}).json()["data"]
if d["state"] == "completed":
print(d["resultUrls"][0]); break
if d["state"] == "failed":
raise SystemExit(d["failMsg"])
time.sleep(3)FLUX 3 Max is a video generation model in the FLUX family by Black Forest Labs: a post-trained, latency-optimized variant of FLUX 3 Video with stronger prompt adherence and better visual quality. It generates 5 to 20 second clips at 720p or 1080p, with optional audio, from text, from a starting image with an optional ending image, or from up to 10 keyframes. On apimodels.app it is the model id flux-3-max.
On apimodels.app FLUX 3 Max costs $0.221 per second of video at 720p and $0.377 per second at 1080p, the same for text-to-video, image-to-video and keyframes. A 5-second 720p clip is $1.105 and a 10-second 1080p clip is $3.77. Until 00:00 UTC on 7 November 2026 a special price of $0.085 per second (720p) and $0.145 per second (1080p) applies. Only successful generations are charged.
Create an API key at apimodels.app, then POST to https://api.apimodels.app/v1/video/generations with "model": "flux-3-max" and a prompt. Add first_frame_url for image-to-video or a keyframes array for keyframes-to-video. The response contains a taskId; poll GET /v1/video/generations?task_id=... until state is completed and download data.resultUrls[0], or pass callback_url to receive the result automatically.
Up to 10 keyframes per clip. Each one is {image_url, frame_index}, where frame_index is a frame number in the 24 fps video (0 to duration × 24), and every position must be unique. On apimodels.app you can write {image_url, time} in seconds instead. Keyframes cannot be combined with first_frame_url or last_frame_url; put the first and last images into the keyframes array at 0 and the clip length.
Yes. Set resolution to 1080p (720p is the default). Audio is generated by default to match the scene; set generate_audio to false for a silent clip, for example when you will add your own soundtrack. Audio does not change the price.
In our spot checks through the apimodels.app API on 20 October 2026, a 720p 5-second clip was ready in 9 seconds for text-to-video, 15 seconds for image-to-video and 19 seconds for keyframes-to-video, with reasoning_level set to low. The publisher reports a 10-second video in about 11 seconds in its own benchmark. Times vary with settings and load; reasoning_level high takes longer.
It depends on the job. FLUX 3 Max is much faster (the publisher reports about 11 seconds for a 10-second video versus 277.8 seconds for Seedance 2.5), goes to 1080p and supports up to 10 keyframes. Seedance 2.5 handles longer clips (up to 30 seconds), up to 50 reference assets and video editing or extension, at 480p or 720p. On apimodels.app at 720p, FLUX 3 Max is $0.221/s at its regular price and Seedance 2.5 is $0.270/s; Seedance 2.5 at 480p is $0.120/s. Both are available with the same apimodels.app key.
Yes, through apimodels.app: you can top up with Alipay, no overseas credit card is needed, and you do not need a separate account with the model publisher. Results come back as direct download links.
Yes. On apimodels.app one API key and one balance cover FLUX 3 Max, Seedance 2.5 and 2.0, Kling V3, Veo 3.1, MiniMax H3 and 140+ other image, video, audio and language models. Switching models is a change of the model field on the same endpoint.
