Image and video models — when to use which

Updated 2026-11-09 · image · Translation: VI

This ranking is a decision board, not a beauty contest. It separates image jobs from short video jobs and scores models on control, consistency, and fitness for shipping assets. “Cool demo” quality without reusable control ranks lower.

At a glance

Common image-model jobs: concept, product, diagram.

Common image-model jobs: concept, product, diagram.

Common video-model jobs: B-roll, avatar, motion.

Common video-model jobs: B-roll, avatar, motion.

Still vs motion: pick the medium that fits the job.

Still vs motion: pick the medium that fits the job.

Criteria (explicit)

Criterion	Image meaning	Video meaning
Controllability	Aspect ratio, references, layout adherence	Motion control, length, camera/path stability
Consistency	Character/product identity across a set	Temporal coherence; fewer morphing artifacts
Text & UI honesty	Readable glyphs; less fake UI	Prefer real capture for instructional UI
Iteration speed	Time-to-usable draft	Time-to-usable clip including regen cost
Production fit	Licensing/terms, resolution, editability	Export practicality for short-form pipelines

Scoring: each criterion 1–5; total shown as /10 after equal weighting (sum of five scores, scaled). Editorial scorecard based on production workflows (thumbnails, concept frames, short social/explainers)—not cinema VFX.

When to use which (quick rules)

Need a precise UI tutorial? Prefer screen recording. Use generative video only for intros or abstract cutaways.
Need a thumbnail or hero still? Use an image model with strong layout control; add text in a design tool if glyphs are weak.
Need five on-brand variants? Prioritize consistency and references over single-image wow.
Need a 2–5s stinger? Short generative video is fine; keep brand marks and fine text out of the generate step when possible.
Need dialogue-heavy explainer? Script + VO + editor beats generative lip-sync for clarity and revision speed.

Image model table (2026-11-09)

Rank	Model tier	Score /10	Use when	Avoid when
1	Top photoreal / design-oriented image models (e.g. current Flagship “pro” image tiers)	8.4	Marketing stills, product-ish scenes, tight prompts with references	You need pixel-perfect UI chrome
2	Strong all-round image models in major chat suites	7.8	Fast ideation inside an existing chat workflow	Long brand series with strict identity locks
3	Stylized / illustration-specialized models	7.5	Consistent illustration systems, gamey/art direction	Photoreal catalog shots
4	Fast cheap image endpoints	6.2	Moodboards, throwaway variants	Final assets with legal/brand scrutiny
5	Local open image stacks	6.0	Privacy-sensitive experimentation, custom LoRAs	Zero-ops teams needing instant reliability

Video model table (2026-11-09)

Rank	Model tier	Score /10	Use when	Avoid when
1	Short-clip generative video flagships	7.6	2–8s B-roll, transitions, abstract visuals	Step-by-step product instruction
2	Image-to-video adapters	7.0	Animating a locked key still	Long narrative continuity
3	Suite-native video tools inside editors	6.8	Light gen inside an editing timeline	Maximum fidelity demos
4	Experimental long-form generators	5.5	Research and prototyping	Client-facing deadlines

Recommended pairings

Job	Primary	Support
Thumbnail	Flagship image	Design tool for final text
Concept frames	Chat-suite image or flagship	Human pick → refine with references
Social stinger	Short-clip video flagship	Still logo animation as fallback
Tutorial short	Screen capture + editor	Generative B-roll ≤15% of runtime
Private brainstorming	Local open image stack	Do not ship without rights review

Anti-criteria

We do not rank higher for:

Fake mobile UIs that mislead viewers
Watermark-heavy free tiers that block publishing
One perfect frame that cannot be reproduced for a series

Changelog

2026-11-09 — Initial board. Category stored as image with explicit video subsection. Next update should pin named SKUs tested and note pricing tier used.