DEV Community: Maya Chen

One browser tab for the still and the clip

Maya Chen — Mon, 19 Oct 2026 08:56:58 +0000

I keep losing the shot when the still lives in one product and the clip lives in another.

VideoAny is a freemium workspace for AI video and image generation and editing. The homepage is https://videoany.ai/. You open it in the browser, make a still or a short clip, download the file, and leave. A paid plan is optional.

It is not a full edit bay. Faces still drift on hard shots, and a long sequence still needs a person. What it gets right is keeping both jobs in one tab so you can see the frame before you spend credits on a clip.

Listing contact: support@videoany.ai. SeekOrigin LLC.

Treat AI video prompts like a spec, not a wish: a template I use for multi-shot clips

Maya Chen — Fri, 25 Sep 2026 16:19:45 +0000

Most bad AI video I have made came from a prompt that read like a mood board: "cinematic, moody, beautiful woman walking in rain, 4K". The model fills every gap with its own defaults, and the defaults change from shot to shot. The fix that worked for me was boring: write the prompt the way you would write a ticket for another engineer.

The five fields

I keep every shot in a tiny block with five fields, always in the same order:

SUBJECT:   who/what, with 2-3 fixed identifiers (hair, jacket color, prop)
ACTION:    one verb phrase, one motion only
CAMERA:    lens feel + one move (slow push-in, static, orbit left)
LIGHT:     source + time of day
CONSTRAINT: what must NOT change (face, logo, outfit)

The CONSTRAINT line is the one people skip, and it is the one that saves the most regenerations. If the jacket is red in shot 1, say "jacket stays red, no pattern" in every shot.

One motion per shot

Asking for "she turns, smiles, picks up the cup and walks to the window" in five seconds is asking for rubber limbs. Split it. Two short clips with one clean motion each cut together better than one clip with four half-finished motions.

Lock the ends, let the middle move

If your generator supports first-frame and last-frame input, use it for anything with a product or a logo. You control the two states that matter (closed box, open box) and the model only has to invent the transition. It is the closest thing to a keyframe you get.

Test harness, not vibes

I run each spec three times before judging it. If one out of three is good, the prompt is not good yet; it got lucky. I only change one field between runs, like a normal A/B test.

Where I run this

Lately I have been testing these specs on Kling4Pro, an independent web front end (not the official Kuaishou site) that exposes Kling 4.0 Pro alongside older Kling versions, Veo 3.1 and Seedance. The useful part for this workflow is that it takes first/last frames and up to nine reference images, and new accounts get some free credits, so you can run the three-run test before spending anything. Whatever tool you use, the spec format is the part that transfers.

If you try the template, the one habit I would keep is the constraint line. It turns "why did the face change" from a mystery into a bug report.

Most AI video tools sell you a clip. Character consistency is the part they dodge

Maya Chen — Sun, 20 Sep 2026 07:51:10 +0000

AI video demos are getting better at the part everyone can see: the first four seconds.

A prompt goes in. A cinematic shot comes out. The lighting is moody, the camera move is expensive-looking, and nobody in the comments asks what happens in shot five.

That is where the demo usually ends. It should not.

The uncomfortable truth is that most AI video tools are selling you a clip, not a character. They can make a person look convincing once. Keeping that same person recognizable while the camera changes, the emotion changes, the outfit changes, and the scene gets longer is a different problem entirely.

The face is not the character

When filmmakers say “consistency,” they do not mean a face that is vaguely similar. They mean the same person has the same identity across a sequence. The jaw should not quietly change shape. A scar should not migrate. Hair should not become a different haircut because the prompt got ambitious. A character’s age, proportions

More motion is not more storytelling

There is also a weird arms race around motion. Every tool wants to prove it can orbit a subject, crash through a window, or perform a drone shot over a city. Fine. But a spectacular camera move around a character who changes identity halfway through is not filmmaking. It is a continuity error with a soundtrack.

The boring work matters more: preserving a reference, extending an action without resetting the actor, and making transitions feel intentional instead of stitched together. If the character is running, the next segment needs to inherit the direction, wardrobe, lighting, and physical state of the previous one. Otherwise the “story” is just a playlist of attractive accidents.

That is why video extension deserves more attention than another prompt-to-clip leaderboard. Tools such as VideoAny are at least pointing at the real bottleneck: what happens after the first generation? Their video extend workflow is the kind of capability I want to judge—not because extending footage magically solves consistency, but because it exposes whether the system can maintain context instead of starting over.

The test I wish every tool published

Forget the cherry-picked hero shot. Give me this test:

Generate a character from a

And yes, price matters. Creators should look at the pricing page, but not just count credits. Ask what a usable minute actually costs after retries, discarded generations, and the time spent repairing continuity. A cheap clip that cannot join the next clip is not cheap. It is inventory you cannot ship.

Stop grading AI video like a poster

AI video will mature when we stop grading it like a poster and start grading it like footage. The question is not “Can it make something impressive?” Almost every serious tool can do that now.

The question is: can I direct a character instead of repeatedly gambling on one?

That is the line between a toy and a production tool. Until vendors publish consistency tests, creators should be skeptical of every perfect six-second example—and especially skeptical of the tools that never show what happens next. reference.

Put them in a scene with a clear action.
Extend that scene three times.
Change the camera angle and emotional beat.
Compare the final frame with the first one.

Show the failures too. Show the prompt, the settings, and the number of retries. If a tool can do that reliably, it has a product. If it can only deliver a beautiful opening shot, it has a demo., posture, and visual history should survive the cut.

That sounds obvious, but the current AI video market rewards the opposite behavior. A glossy single clip is easy to screenshot and post. A coherent 45-second scene is harder to generate, harder to evaluate, and much harder to fake with a marketing page.

So the industry keeps optimizing the trailer instead of the workflow.

Paying for AI video that melts faces is a tax on optimism

Maya Chen — Sat, 19 Sep 2026 14:00:19 +0000

I keep meeting marketers who renew a "pro" AI video plan out of guilt, not results. Pitch says cinematic; export looks like three cousins sharing one hoodie.

Continuity isn't optional. Seedance 3.0: https://seedance30.com/seedance-3-video-generator
Multimodal: https://seedance30.com/seedance-3-multimodal-video
Compare: https://seedance30.com/veo-3-1-lite-vs-kling-3-0-pro
Pricing: https://seedance30.com/pricing
Home: https://seedance30.com/

Stop funding the costume-change economy.

AI video vendors sold us "native audio." Most meant "export silent MP4"

Maya Chen — Sat, 19 Sep 2026 11:14:50 +0000

Hot take from someone who edits for money: silent AI clips with a CapCut soundtrack are not a product category. They're a half-finished export button with a marketing budget.

I wasted a quarter testing tools that promised lip sync and delivered mouth flaps that looked like bad puppet theater. Clients notice. Viewers notice. Algorithms notice when people bounce at second three.

What changed my mind wasn't another polished landing page. It was running Seedance 3.0 with uploaded audio guiding the visuals — see https://seedance30.com/seedance-3-video-generator — and getting sync that was "good enough to ship" without a second app. Still weird on hard consonants sometimes. Still better than the silent-file tax.

If you're stuck in comparison hell, this page is blunt: https://seedance30.com/veo-3-1-lite-vs-kling-3-0-pro

Credits/pricing without the soft sell: https://seedance30.com/pricing

Root: https://seedance30.com/

Multimodal path if you want image+audio+text in one flow: https://seedance30.com/seedance-3-multimodal-video

I'm not saying fire your editor. I'm saying stop buying tools that force you to rebuild the soundtrack from scratch every time the model "improves."