When to Fall Back to Seedance 1.5 Pro

Seedance 2.0 is better at almost everything than its predecessor. Arena Elo 1270 for text-to-video against 1173 for 1.5 Pro. Tighter motion coherence. Cleaner identity holds. But 2.0 caps at 720p, and 1.5 Pro still ships 1080p natively. For a meaningful slice of production work, that spec flips the choice back to the older model.

The resolution cap is the whole story

Seedance 2.0 outputs a maximum of 720p. If your delivery contract specifies 1080p master files, you cannot hand in 720p upscaled after the fact. Broadcast partners and large brand clients will reject it.

Seedance 1.5 Pro, at fal-ai/bytedance/seedance/v1.5/pro/text-to-video, still outputs 1080p directly. Lower Arena Elo, 1173 versus 1270: a real quality gap of roughly 100 points that translates to softer motion coherence and weaker identity holds. But 1080p native beats 720p upscaled in a delivery package.

Case one: broadcast deliverables

A broadcast spot on cable or OTT needs 1920 by 1080 minimum. The spec sheet is non-negotiable. Generate at 720p and upscale in post and you add a soft pass every colorist will notice. Grain goes wrong, small text melts, logos feather.

Use 1.5 Pro for broadcast. The Elo gap shows up in motion, not in per-frame image quality. A well prompted 1.5 Pro shot at 1080p native looks cleaner in a broadcast grade than a 720p 2.0 shot pushed through a spatial upscaler.

01import { fal } from "@fal-ai/client";
02
03fal.config({ credentials: process.env.FAL_KEY });
04
05const result = await fal.subscribe(
06  "fal-ai/bytedance/seedance/v1.5/pro/text-to-video",
07  {
08    input: {
09      prompt: "Wide establishing shot, empty arena at dusk, stadium lights cycling on, slow push in toward center ice.",
10      duration: 5,
11      resolution: "1080p",
12      aspect_ratio: "16:9"
13    }
14  }
15);
16
17console.log(result.data.video.url);

1.5 Pro does not support the twelve reference stack that 2.0 does. You get prompt and an optional image reference. Plan accordingly.

Case two: upscale pipelines that need headroom

Upscaling to 4K for a cinema screen or digital billboard works better with more starting pixels. 1080p to 4K is a 2x linear ratio. 720p to 4K is 3x. The upscaler is inventing more, and more invention means more artifacts.

For any pipeline that ends at 2K or above, start with 1.5 Pro at 1080p.

Case three: archival clients requiring 1080p master

Some clients in legal or compliance sensitive industries contract for 1080p masters with specific codec requirements. Banks, pharma, enterprise training content. The spec is in the statement of work because someone in legal wrote it ten years ago and has not updated it.

You cannot deliver a 720p master against that spec without a contract amendment. Run 1.5 Pro for those projects.

When 2.0 still wins

Most social work. A 9:16 feed on a phone is fine at 720p. The Elo advantage of 2.0 shows up in motion quality, which phone viewers notice.

Reference stacks. 2.0 accepts up to twelve references across images, video, and audio. 1.5 Pro does not.

Audio-driven output. 2.0 lip-syncs and rhythm-matches. 1.5 Pro does not ship with that capability.

A decision table for producers

Deliverable is 720p or below, final destination is social: use 2.0. Deliverable is 1080p master, no upscale planned: use 1.5 Pro. Deliverable is 4K after upscale: use 1.5 Pro, then upscale. Brief has four or more references: use 2.0. Brief needs lip sync: use 2.0. Broadcast or OTT contract with strict spec: use 1.5 Pro unless 720p is permitted.

Parallel submission when unsure

When you are unsure which model wins for a shot, submit both and compare. Cheap insurance on a spec-sensitive project.

01const [v2, v15] = await Promise.all([
02  fal.subscribe("bytedance/seedance-2.0/text-to-video", {
03    input: { prompt, duration: 5, resolution: "720p" }
04  }),
05  fal.subscribe("fal-ai/bytedance/seedance/v1.5/pro/text-to-video", {
06    input: { prompt, duration: 5, resolution: "1080p" }
07  })
08]);

Review side by side. You will know within thirty seconds which to deliver. The 100 point Elo gap matters but is not a chasm. A well prompted 1.5 Pro shot at 1080p looks like a working render from a good model. Where the 720p cap on 2.0 disqualifies it, 1.5 Pro is the right fallback.