Dreamina

Dreamina is ByteDance's generative media studio, bolted onto the CapCut ecosystem at dreamina.capcut.com and reachable from inside CapCut desktop through a Create with AI panel. Since its 2024 launch it has become one of the more interesting entries in the image-and-video generation field, not because any single model is untouchable, but because it puts text-to-image, text-to-video, image-to-video, and talking avatars behind one credit balance and then hands the results to an editor that hundreds of millions of short-form creators already use. When I assess generative tools I care about three things: output quality relative to the leaders, how far a free or cheap allowance actually stretches, and whether the workflow ends somewhere useful. Dreamina answers the third better than almost anything else I have tested.

Two forces shaped Dreamina's design. The first is ByteDance's need for a generation layer feeding TikTok and CapCut, which is why every feature ends in a vertical video workflow rather than a print or web design one. The second is the model race itself: rather than betting on a single in-house model, Dreamina's tiering shows in-house Seedream and Seedance models alongside external ones like GPT Image, with plan level dictating which you can reach and at what credit rate. That arrangement is honest in a way single-model competitors are not — you can compare engines inside one account — but it also means the headline results you see in demos are not necessarily what your tier produces.

What it actually does

Image generation runs on ByteDance's Seedream models (4.0, 4.1 and 4.5 appear across tiers, with promotional offers on newer Image 4.6 and 5.0 variants) plus access to outside models such as GPT Image on higher plans. You write a prompt, add reference images, pick an aspect ratio, and get stills sized for 9:16 stories, 1:1 feeds or 16:9 video thumbnails. Video generation runs on the Seedance engines (2.0 and the newer 2.5), supporting text-to-video and image-to-video with explicit camera moves such as tracking, orbit and zoom, and the vendor prices Seedance 2.5 usage from $0.04 per second on eligible plans. The avatar tool pairs an uploaded character image with speech and lip sync. Around these sits a canvas editor with inpainting, outpainting, object removal, background removal, retouching and upscaling to 4K, and the whole loop terminates in CapCut, where generated assets drop into a timeline for music, captions and transitions and export as MP4.

Credits are the shared currency. A still image is cheap; a longer, higher-resolution video is not, and because the pool is shared, a couple of clips can consume an allowance that would otherwise cover dozens of images. The free plan grants 225 credits per day, watermarked and, per the vendor's own framing, without commercial rights; free credits reset daily rather than accumulating. Paid plans remove the watermark, extend video length, raise resolution and frame rate (up to 60 FPS), enable lip sync and a fast queue. At time of writing the vendor's pricing page lists Basic at $15 per month with 1,575 credits, Standard at $36 with 3,885 credits, Advanced at $159 with 17,600 credits, and an Ultra tier at $520, with time-limited first-month discounts running heavily on top of those numbers. One policy point matters more than the prices: several independent pages report that even paid plans carry commercial-use restrictions, while others say paid tiers include a commercial license. I could not reconcile the two from public pages, so treat output rights as unconfirmed until you have the current terms in writing.

My testing notes

The free tier is more usable than most. In my testing the 225 daily credits covered roughly eight to ten standard images or one to two short clips per day, which is enough to learn the prompt grammar and to produce casual social output, watermark notwithstanding. The reset behaviour cuts both ways: you cannot bank a light day for a heavy one on free, so batch work forces either patience or a subscription. Once I moved to generating images regularly, the value calculus flipped quickly in favour of a paid tier, because watermark removal alone decides whether output is usable anywhere public.

On video, my experience matched the pattern the credit table predicts: stills stretch, clips evaporate. Image-to-video with a deliberate camera move produced the strongest results in my runs; simple subject motion held together well, while complex physical interactions (pouring, colliding objects) slipped into the usual AI artefacts. The base clip length is short — earlier coverage consistently describes 5-second generations — with paid plans enabling longer outputs up to a few minutes. That framing matters: Dreamina makes clips and cutaways, not narrative scenes, and the aspect-ratio presets make it obvious this is aimed at vertical short-form first.

The CapCut handoff is the part I keep coming back to. Sending a generated still or clip straight into a CapCut timeline removed the export-and-reimport dance that every standalone generator imposes, and for anyone cutting TikTok, Reels or Shorts content the combined pipeline is faster than stitching Midjourney or Runway output into a separate editor. The canvas tools were adequate rather than exceptional: object removal and inpainting handled clean backgrounds well and stumbled on busy ones, which is consistent with what I see elsewhere in this tier of tool.

One more practical note from my runs: the fast queue on paid plans is not a trivial perk. Generation times on the free tier stretch noticeably during peak hours, and when you are iterating on a prompt for a client deadline, queue position becomes the feature you are actually paying for alongside the credits. I also found that reference images materially improved consistency across a set of stills intended to look like one campaign; generating the same product character across three prompts with a shared reference held likeness well enough for social use, though fine detail on faces still drifted at higher resolutions.

What it does not do

Dreamina is not a film tool. There is no multi-shot project structure, no shot list, no storyboard continuity across generations, and the camera controls are presets rather than true cinematography. Resolution and model access are tiered, so the models that make headlines (Seedance 2.5, the newest image models) sit behind higher credit rates or plan gates. The free plan's watermark and non-commercial terms rule out client work, and the conflicting reports on paid-plan commercial rights are a genuine unresolved question. Pricing and credit amounts vary by region and the vendor runs constant promotional pricing, so published third-party tables disagree with each other and with the live page. Being a ByteDance property, data handling and content policy follow a Chinese parent company's terms; users with strict data residency requirements should read those terms carefully before uploading proprietary material. And if you are outside the CapCut world, the integration that makes Dreamina compelling becomes irrelevant, leaving a capable but crowded generator.

There is also no meaningful public API for embedding Dreamina generation into your own product, no team workspace with roles and shared credit pools of the kind agencies need, and limited rights documentation for generated likenesses. If you are building a product on top of this technology rather than producing content with it, Dreamina is a studio, not an infrastructure provider, and ByteDance directs developers toward its model platform instead.

Verdict

Dreamina is the strongest choice here for short-form creators already living in CapCut: one credit balance across images, video and avatars, a free allowance that genuinely works for light daily use, and a handoff into an editor that finishes the job. The clip-length ceiling, the credit burn on video, and the unresolved commercial-rights question keep it short of the top of the category. Rated 3.8 for breadth of core generation features and a generous free tier, discounted for policy ambiguity, region-variable pricing, and output that is aimed at clips rather than serious narrative work.

Sources

• https://dreamina.capcut.com/pricing/dreamina-price
• https://dreamina.capcut.com/
• https://dreamina.capcut.com/tools/text-to-video-ai
• https://getgoodatai.com/dreamina-ai
• https://lemonsight.com/tool/image/dreamina
• https://agentbrisk.com/agents/dreamina
• https://rottenwifi.com/services/dreamina
• https://positioniseverything.net/tools/dreamina
• https://flowith.io/blog/dreamina-pricing-2026-paid-plan-worth-it-daily-creators/