Starter
Around 5 clips
$9.90one-time
5clips
$1.78 per clipBase price
1,000 credits
Two video models with sound. FLUX 3 Max is the fast and cheap one; Seedance 2.5 is ByteDance's longer, more controllable one. Every figure here is published by someone else and sourced at the bottom; we made no head-to-head test. We sell FLUX 3 Max only.
FLUX 3 Max: 180 credits for a 5-second clip at 720p. Failed videos are free.
A failed video is refunded.
Speed and price favour FLUX 3 Max. fal's own benchmark has a 10-second 720p clip in 12.45 seconds on Max against 277.80 on Seedance 2.5, about 22 times faster, and fal lists Max at $0.17 a second at 720p against about $0.47 for Seedance 2.5.
Reach and control favour Seedance 2.5. ByteDance says it makes 30-second clips in a single pass, against 20 seconds for Max, and takes up to 30 images, 10 video clips and 10 audio clips as references in one pass, where Max takes a first and an optional last frame (or keyframes through fal's API).
Quality depends on the board you read. Artificial Analysis' text-to-video board has the two level: FLUX 3 Max at 1157 and Seedance 2.5 at 1156, intervals overlapping. On its image-to-video board Seedance 2.5 is at the top (1155, rank 1 to 2) and FLUX 3 Max is lower (1065, rank 11 to 13). fal's own launch page shows both at 1,139 on an earlier preliminary snapshot of that board; the current board no longer shows that tie.
Figures as published by the sources at the bottom. Elo values are Artificial Analysis' human-preference scores with 95 percent intervals, read on 3 November 2026.
| FLUX 3 Max | Seedance 2.5 | |
|---|---|---|
| Made by | fal Research, post-trained with Black Forest Labs on FLUX 3 | ByteDance |
| Released | October 2026 | 31 July 2026 |
| Length | 5 to 20 seconds | 4 to 30 seconds in one pass on fal; ByteDance also describes extending in several rounds |
| Resolution on fal | 720p; 1080p as an upscale | 480p and 720p in the API |
| Inputs | Text; a first and an optional last frame; keyframes | Text; an image with end-frame control; reference-to-video with up to 30 images, 10 video clips and 10 audio clips (50 files in all) |
| Audio | Generated with the video; can be turned off | Generated with the video; dialogue in quotation marks is spoken and lip-synced, says fal |
| Price a second on fal | 720p $0.17, 1080p $0.29 (promotion until 7 Nov 2026: $0.0425 and $0.0725) | 480p about $0.22, 720p about $0.47 (billed by tokens) |
| Time for a 10-second 720p clip (fal's benchmark) | 12.45 seconds | 277.80 seconds |
| Image-to-video Elo (Artificial Analysis) | 1065 (1049 to 1081), rank 11 to 13 | 1155 (1139 to 1171), rank 1 to 2 |
| Text-to-video Elo (Artificial Analysis) | 1157 (1140 to 1174), rank 2 to 8 | 1156 (1142 to 1170), rank 2 to 7 |
| Price listed by Artificial Analysis | $4.35 per minute of video | $34.12 per minute of video |
A 10-second clip at 720p is 360 credits here, $3.00 to $3.56. On fal it is $1.70 at the regular price against about $4.73 for Seedance 2.5, and it comes back in seconds instead of minutes.
Many reference images, video and audio clips in one generation, and clips up to 30 seconds. If your brief is a storyboard with a cast, that is the one built for it.
On Artificial Analysis' current image-to-video board it is first and FLUX 3 Max is not close. If the photo is the point and quality matters more than time, look there first.
Short clips and many takes are where the speed and the price per second matter most. The text-to-video board has the two level.
Three clips from 3 November 2026: 15 seconds, 20 seconds and 1080p. We did not make Seedance clips, so there is nothing to put beside them; judge each on its own.
Skip it if you need a clip longer than 20 seconds in one go, a pile of reference images, videos or audio files, or lip-synced speech from a script: those are Seedance 2.5's strengths on the sources we read. Skip it for 4K too: Max does not make it.
And do not choose it because of the launch claim that it ties for first on image-to-video: on Artificial Analysis' current board it does not. Choose it because it is quick, because a 5-second clip costs 180 credits here, and because the text-to-video board has it level with the leaders.
A 5-second clip at 720p costs 180 credits (36 a second); 1080p costs 62 a second. Buy a pack once and use it whenever you like. Bigger packs bring the price of a clip down from $1.78 to $1.50.
Starter
$9.90one-time
5clips
$1.78 per clipBase price
1,000 credits
Basic
$19.90one-time
11clips
$1.71 per clipSave 4%
2,100 credits
PlusRecommended
$39.90one-time
24clips
$1.63 per clipSave 8%
4,400 credits
Pro
$79.90one-time
51clips
$1.56 per clipSave 12%
9,200 credits
Studio
$149.90one-time
100clips
$1.50 per clipSave 16%
18,000 credits
Secure one-time checkout through Stripe. Buying a pack only adds credits to your account. USD prices. Review the final amount and any applicable tax at checkout before paying. Credits do not automatically renew, and unused credits can be refunded within 7 days of purchase. The credit cost shows before each video; finished results need no second payment to view or download.
Every outside figure on this page, and where we read it. We did not run any of these models ourselves for this page.
| Source | What we took from it | Read on |
|---|---|---|
| fal.ai/flux-3-max (updated as of 18 Oct 2026) | 12.45 s and 277.80 s for a 10-second 720p clip; both at 1,139 Elo on Artificial Analysis' preliminary v1.1 image-to-video snapshot; Max $0.0425 a second at 720p with list $0.17 | 3 Nov 2026 |
| fal.ai/models/blackforestlabs/flux-3-max/text-to-video | $0.17 and $0.29 a second, the promotion to 7 Nov, 5 to 20 seconds, 720p with optional 1080p upscaling | 3 Nov 2026 |
| fal.ai/models/bytedance/seedance-2.5/text-to-video | About $0.2205 and $0.4730 a second at 480p and 720p, token billing, 4 to 30 seconds, 480p and 720p in the API, audio on by default, up to 50 reference inputs on the reference endpoint | 3 Nov 2026 |
| fal.ai/learn/devs/seedance-2-5-vs-seedance-2-0 (published 18 Aug 2026) | Dialogue in double quotes voiced and lip-synced; 4K not on fal | 3 Nov 2026 |
| seed.bytedance.com, Introducing Seedance 2.5 (31 Jul 2026) | Up to 30 seconds per generation; up to 30 images, 10 video clips and 10 audio clips as references; ByteDance's own note that complex motion plausibility can improve | 3 Nov 2026 |
| artificialanalysis.ai/video/leaderboard/image-to-video (AA-Video-I2V v2.1) | Elo, intervals, ranks, release months and price per minute for both models | 3 Nov 2026 |
| artificialanalysis.ai/video/leaderboard/text-to-video (AA-Video-T2V v2.1) | The same for the text-to-video board | 3 Nov 2026 |
fal sells both models, so its speed figures are its own benchmark. Artificial Analysis' boards change as votes arrive; read the live pages for today's numbers.
It depends on what you need. FLUX 3 Max is much faster and cheaper per second. Seedance 2.5 makes longer clips with more references and ranks higher on Artificial Analysis' image-to-video board. On the text-to-video board the two are level.
FLUX 3 Max, by a wide margin on fal's own benchmark: 12.45 seconds against 277.80 for a 10-second 720p clip, about 22 times. Real times vary with load; our own 5-second Max clips took 10 to 34 seconds.
On fal, Seedance 2.5 is billed by tokens and works out at about $0.47 a second at 720p; FLUX 3 Max is $0.17 a second at 720p, with a 75 percent promotion until 7 November 2026.
FLUX 3 Max: 5 to 20 seconds. Seedance 2.5: up to 30 seconds in a single pass, according to ByteDance and fal.
Yes. Both generate audio together with the picture and can be used with sound off. fal says Seedance 2.5 voices and lip-syncs dialogue written in quotation marks.
No. We offer FLUX 3 Max only.
fal's launch page quotes an earlier preliminary image-to-video snapshot, where both were at 1,139. The current board, read on 3 November 2026, has Seedance 2.5 at 1155 and FLUX 3 Max at 1065.
Flux3Max.io is an independent service. It is not affiliated with, endorsed by or sponsored by Black Forest Labs or fal; FLUX is a trademark of Black Forest Labs, and model names belong to their owners.
One account and one set of credits for every page.
New accounts get a free first clip. A failed video costs nothing.