I Tried Adding AI Video To My Brand Identity Service. Here's What Stuck.

For most of the last decade I sold the same package. Logo system, type pairing, color guide, a few mockups, a PDF brand book, done. Repeatable, predictable, capped at a number my market would pay. Every couple of years I'd raise rates a little and lose maybe one in five prospects on price.

Two things kept eating at me. First, the engagement always ended too soon. You spend six weeks getting deep into a brand's voice, then hand off a static system and the relationship goes cold. Second, the clients who really moved the needle post-launch were the ones with video to plug into their channels. About-page reels mostly, plus vertical content for socials. The brand systems I was building got more juice in the hands of clients who had also paid someone else to make the video, which was a slightly humiliating realization to sit with.

I'd looked at folding motion into my deliverables a few times. Every time the math killed it. After Effects in-house meant losing three days of my week to keyframing I hadn't done seriously in years. Subbing out to a motion designer meant either passing the cost to the client (sticker shock) or eating the margin (no thanks). The freelance brand identity tier just doesn't have room for a $2k motion sub.

Then I started testing AI video models for production work instead of doomscrolling demos on X. The model that didn't fall apart at brand-adjacent prompts, and is now part of how I scope new engagements, is ByteDance's Seedance 2.0.

I'll save you the eight other models I tried first.

The fastest way to see if it does what I'm claiming is to throw a brand prompt at it directly. seedance2.so runs the model with no signup hoops, and the first three credits are free.

What Won Me Over After Eight Other Models

The short version: it generates fast enough that I can iterate on prompts the way I iterate on type choices. That's the whole game.

The longer version involves specs. ByteDance ships two variants. The Standard one (doubao-seedance-2-0-260128 if you go through the API directly) is the quality tier. The Fast one (doubao-seedance-2-0-fast-260128) trades a small amount of fidelity for shorter render times, and Fast is where I burn most of my credits. I run twenty variants quickly, throw away eighteen, regenerate the two survivors at Standard for the deliverable.

Clip length sits anywhere from four to fifteen seconds, which is the range I need. Resolution caps at 720p natively. No 1080p out of the box. I had concerns about this on paper. After producing maybe forty clips for client work, the concern never materialized. The deliverables go on About pages and LinkedIn cover slots, sometimes into pitch decks viewed on screenshare calls. Neither environment cares about the difference between 720p and 1080p. If you're delivering for a Vimeo film festival you should be looking somewhere else anyway.

The audio generates alongside the video in the same call. This sounds like a small detail. It's the thing that actually changed my workflow. The previous mental model was: generate video, then sync to a music bed as a separate step. Seedance bakes ambient audio into the output, so the deliverable arrives with a usable rough sonic layer that the client either keeps or swaps. Often they keep it.

Then there's omni-reference, which is the workhorse mode for brand work. You can pass in up to nine reference images, three reference videos, and three reference audio clips per call. For a brand engagement I drop in: the wordmark, the symbol mark, the brand color reference image I already made for the static deliverables, two or three "vibe" stills that informed the mood, and a snippet of the brand's existing music if they have one. The output stays visually within shouting distance of the brand. It isn't magic and it doesn't replace art direction. It does produce variants that look like they belong to the brand instead of generic AI sludge, which is the bar.

There's a service tier called "flex" that runs the same job offline at half the cost. I queue Friday batches through it and the renders land in my inbox by Monday morning. Useless for live client meetings. Excellent for buffer stock between revision rounds.

How I Repackaged The Deliverables

Adding "we can do motion" to a brand engagement doesn't change the pricing math by itself. What changes the math is breaking the motion deliverables into named, separately-priced add-ons. Four of these earned their place in my current proposals.

The brand anthem reel

A 12-15 second mood film that captures the feeling of the brand. Goes on the About page. Plays at the front of pitch decks. The work is mostly in writing the prompt right (what does this brand feel like in motion?) and generating fifteen variants with the omni-reference assets loaded. Pick two finalists. Four hours including curation. I price it as a half-day add-on and clients say yes more often than not.

The animated logo reveal

The interpolation mode (first frame and last frame, the model fills in the motion between) handles this. Give it a starting frame of the wordmark fading in, an ending frame of the full logo locked on the brand color background. Generate eight variants. Two will be wrong in a way that's funny. One will be the one you ship. Bill it as a flat add-on.

The social motion drop

The kit-style deliverable. Twelve to fifteen vertical clips generated as a batch through the flex tier overnight, all anchored to the brand assets via omni-reference. The client uses them as social posts during the brand launch window. I hand them off in a folder with a naming convention that references the brand's content pillars. This is the deliverable that surprises clients most, because nobody else in their RFP shortlist is offering it.

The pitch deck interlude pack

The repeat-business deliverable. Six months after the brand engagement closes, the client is raising a round or making a sales pitch and needs short video sequences to break up the slides. Same omni-reference assets, same brand consistency. New invoice, same engagement.

The Setup Bit

If you're integrating this into a studio with multiple seats and a real pipeline, you go to Volcengine direct. That's where ByteDance hosts the official API. It's also a developer-first environment that assumes you have a business entity and an engineering resource on hand. Most freelancers don't, and don't need to.

For my own work I run everything through seedance2.so, which is a studio interface built on top of the same model. Each generation mode lives on its own page (text-to-video, image-to-video, omni-reference, video extension), parameters appear as sliders and dropdowns, and you never deal with JSON. Three free credits on signup got me through the initial testing without committing to anything. Other third-party wrappers exist. I haven't had a reason to switch once I had this one set up.

The reason I'm spending this many words on the access layer is because the friction of getting started is what kills most designers' interest in adding AI video to their service. Try one of the wrappers with a previous client's brand assets in the omni-reference slot and see what comes back. It either fits the way you work or it doesn't, and the only way to find out is the boring way.