Colossyan 20-30 minute video cap vs Sume avatar videos of 4-60 seconds

Colossyan is built for long training videos and Sume's avatar video is built for short clips, and the limits on their pages show it. Colossyan's pricing page lists a video duration limit of 20 minutes on Starter and 30 minutes (monthly) or 50 minutes (yearly) on Professional, with 40 scenes. Sume's docs accept a script or multi-scene plan only when the estimated duration is 4 to 60 seconds.

What are Colossyan's limits?

The comparison table on the pricing page, read on 2026-10-03, lists video minutes per month as well as per-video limits. Starter shows 20 minutes a month with the NEO avatar model and a little under one minute with NEO2; Professional shows 30 minutes with NEO and 10 minutes with NEO2; Enterprise shows unlimited. The per-video row reads 20 min for Starter, 30 min monthly or 50 min yearly for Professional, and Custom for Enterprise.

What are Sume's limits for an avatar video?

The Avatar 1.0 talking-video route takes a ready avatar plus either a script or ordered video_inputs. The docs say scripts and multi-scene plans are accepted when Sume estimates the target duration at 4 to 60 seconds inclusive, and tell you to shorten longer scripts or split them into multiple jobs. Multi-scene plans use voice.type: "silence" for non-speaking beats, and the planned duration must still land in the same window. Current execution supports one resolved avatar per final video, and the output resolution is currently 720p.

Inline captions are optional, and an estimated duration above 60 seconds is rejected for them.

How do you get a longer video on Sume?

You assemble it. Generate several avatar clips of up to 60 seconds and join them with Timeline 1.0, which takes an audio spine plus an ordered video[] list and returns one MP4 at $0.10 per ceil(output minute). The audio can be built from up to 20 gapless slices with no re-synthesis, and transitions are limited to a small set such as fade and dissolve, each at most one second.

That is more work than a single long render, and the docs make no claim that the pieces will feel like one continuous performance. For a 10-minute training module with many scenes, a long-form editor is the closer fit; for a stream of short, personalized clips, the job-per-clip shape fits an API.

Which limit matters to you?

Count the length of your finished video and the number of them per month. Long internal training with a monthly minutes allowance points to Colossyan's model. Many 30-second clips per day, billed by the job and driven from code, points to the Sume shape. This is a statement about limits only, not about how the avatars look or sound.