Seedance 2.5: fal's 30 s at 720p listing vs Sume's catalog limits
fal.ai/models says Seedance 2.5 is a native 30-second clip at up to 720p. Sume's catalog lists 4 to 30 s at 480p, 720p and 1080p. Read the live values.

fal's model list describes Seedance 2.5 as a native 30-second clip at up to 720p. Sume's video docs list seedance-2.5 at 4 to 30 seconds and 480p, 720p or 1080p. Two pages, two descriptions: the first is a summary line on a listing, the second is the capability data Sume's catalog exposes.
Do not reconcile them by guessing. Ask the catalog you are about to call.
The two descriptions
As read on 2026-10-03.
| Item | fal.ai/models listing | Sume video docs |
|---|---|---|
| Duration | Native 30-second clip | 4 to 30 seconds |
| Resolution | Up to 720p | 480p, 720p, 1080p |
| Kind of statement | One-line listing description | Capability data from GET /v1/videos/models |
| Pricing | Not in the entry reviewed | pricing_skus per model in the catalog |
Why they can differ
A listing line summarizes a model family and may describe one provider's configuration. Sume documents that limits are not uniform across its catalog and tells you to check the models endpoint before submitting. Neither page here is wrong on its face, and neither supports concluding which gateway is better. The only claim that matters for a build is what the endpoint you call accepts today.
Read the live limits
This script lists the video models Sume exposes and prints the fields that matter for a plan: id, supported resolutions and durations. Field names beyond supported_resolutions are read defensively, since the docs describe the catalog as the source of truth and the response can grow.
import os, requests
r = requests.get("https://api.sume.com/v1/videos/models",
headers={"Authorization": "Bearer " + os.environ["SUME_API_KEY"]},
timeout=60)
r.raise_for_status()
for m in r.json()["data"]:
if "seedance" not in m["id"]:
continue
print(m["id"])
print(" resolutions:", m.get("supported_resolutions"))
print(" durations: ", m.get("supported_durations"))Using the result in a request
Once the catalog confirms a value, pin it in the request rather than leaving it to a default.
- Set
model,resolutionand the duration explicitly, within the range the catalog lists for that model. - Use
sume/autowhen you want the router to choose, and pin a model id when a pipeline depends on a specific resolution. - Re-read the catalog on a schedule. Models are retired, and the dates are on vendor pages.
- Treat any third-party listing, including fal's, as a pointer to go and check the catalog, not as the limit.
Sources
Related posts
More in Comparisons
- Stable Diffusion alternatives in 2026: open weights or hosted API
SD 3.5 is still Stability's newest flagship image model. FLUX.2 [klein], Qwen-Image 2.0 and Z-Image Turbo are the open options. How to pick between them.
- Suno Speech beta: voice and music in one pass, or separate tracks?
Suno's Speech beta makes voice and music in one track. Its blog lists wandering accents and long pauses. When to prefer separate TTS, music and a timeline mix.
- Suno v6 and label partners: what it means for an ad soundtrack
Suno's v6 post names WMG, BMG and Believe and upload screening. What a rights-minded team should check before using any AI music in a paid ad.
- Reading a vendor-run TTS leaderboard: Gemini 3.8 Flash TTS on VoiceEQ
Hume's blog lists Gemini 3.8 Flash TTS atop its Real-World VoiceEQ board. What that does and does not tell you, plus a blind test to run on your script.
Written by Sume