108August 30, 2026·3 items

MovieFlow, Arena and Perchance: all three work, all three sample prompts are wrong

The three tools on the list are real and all three run for free. The problem is the sample prompts: all three write an interface setting into the prompt text. Typing "generate 32 variations" or "test across 3 AI models" does nothing, those numbers come from a dropdown. The second issue is the name: the site listed as "Arena" is LMArena's new name since January, and its real value isn't comparing models, it's trying Sora and Veo for free. The third is the word "free": it means three different things across these tools, and in one of them it breaks the faceless channel idea outright.

VideoGörselPrompt

The three tools at a glance

All three links sit at the bottom of this page, each item has its own button. If you only came for the addresses, jump straight there.

Tool	What it does	Where the free tier stops you
MovieFlow	A multi-scene cartoon from a single prompt	The download comes out watermarked
Arena	Runs the same prompt on two models side by side	Anything you type can become research data
Perchance	Up to 32 images in one go	The commercial license is unclear

What's extra below: why all three sample prompts are wasted lines, Arena's real name and the free video side the list never mentions, what you shouldn't type into Arena, and what the word "free" separately means in each of these three tools. None of the five are in the original list.

The mistake all three sample prompts share

In all three of these tools the count, the duration and the model choice are settings that live in the interface. The prompt box doesn't read them. All three examples in the original list wrote those settings out as sentences, which hands the model an instruction it can't act on. The result isn't even an error: the tool either quietly treats that line as part of the story or ignores it entirely. Both work against you.

Typing "generate 32 variations" into Perchance. The prompt box has no concept of count, so the line bleeds into the image content and you still end up with the default 4.

Describe only the image in the prompt and pick 32 from the image count dropdown under the box. The menu offers 2, 4, 6, 8, 16 and 32, and it defaults to 4.

Typing "test across 3 AI models" into Arena. Model choice happens on screen, not inside the prompt, and there's no mode that runs three models at once anyway.

Write only the output you want and pick the mode on screen. Battle races two anonymous models, side by side lets you choose the two yourself.

Writing "make it 2 minutes" into MovieFlow and moving on. Duration, language and aspect ratio are chosen on the settings screen before generation.

Set the duration in the settings and spend the prompt on scene order and tone. Listing scenes works; stating the duration in a sentence doesn't.

The "Arena" on the list is LMArena's new name

The original list describes it as some anonymous "multi-model testing platform", when it's the best known site in the space. It started as Chatbot Arena, moved to lmarena.ai in 2024, then took the name Arena and moved to arena.ai on 28 January 2026. The old address redirects to the new one. So it's the same place you may already know as LMArena.

Knowing the name helps because the site's main job isn't chatting, it's ranking. There are separate leaderboards for text, image, video, image editing, front-end coding and agents, all computed from real user votes. Before you pay for any model you can see where it actually stands that month.

Three modes, three different jobs

Mode	How many models	When to use it
Battle	2, names hidden	Unbiased comparison. Names are only revealed after you vote.
Side by side	2, both picked by you	When you already have two candidates in mind. Votes here don't feed the leaderboard.
Direct chat	1	Using a single model plainly. No voting.

That's why the original list's "run the same prompt across multiple AI models simultaneously" is only half right. The number running at once is two, and which ones you get to choose changes by mode. The sample prompt asking for three models has no counterpart on any of these screens.

The part the list never mentions: trying Sora and Veo for free

Arena's real value for a content creator isn't reading the leaderboard, it's joining the voting. On the video side, the models generate from your own prompt while you vote, which means you get output from frontier video models without ever touching a checkout page. The text-to-video board currently ranks 46 models: Google's Gemini Omni versions, Black Forest Labs' Flux video model, ByteDance Seedance, OpenAI Sora, plus entries from MiniMax and Alibaba.

It works exactly like battle mode: you write the prompt, the system randomly pairs two models, both generate, you pick the better one, and the names appear after the vote. You walk away with two different model outputs and the answer to which was which.

It isn't unlimited, and I'm not putting the number here

Video generation requires an account and comes with a daily cap. The cap is low, a few generations' worth. I'm not writing the exact figure because the only sources I could find are third parties and the platform itself says the limit will move with demand. You'll see the real number on screen. What to plan around: you can't run ten attempts in a day, so sit down and write the prompt properly once.

What not to type into Arena

This site's whole model is built on collecting your data, and it says so itself. Your conversations and certain other information can be disclosed to the relevant model providers, and can also be published in public research datasets. Shared data is de-identified, but the content itself remains.

This is what the free part costs

Client work, an unpublished video idea, a brand brief, a list of names, a product you haven't announced, personal information: none of it goes in here. Arena isn't a scratchpad, it's a public testing ground. If the thing you want to test is confidential, turn it into a generic example first and type that. Opting out of data use means emailing their support; it isn't off by default.

"Free" means three different things here

The original list drops all three into the same bag: free, fast, ideal for content creators. But the free part ends somewhere different in each one, and it ends exactly at the moment you go to publish.

Tool	What's free	What isn't
MovieFlow	Signing up, building the story, previewing	Watermark-free download and commercial use
Arena	Leaderboards, chat, video and image generation	Your privacy, and anything past a few generations a day
Perchance	Everything, there isn't even a sign-up	A clear commercial license and any promise of consistent quality

For anyone building a faceless channel the column that matters isn't the middle one, it's the right one. A channel earning ad revenue counts as commercial use.

Perchance: the 32 is right, the rest is missing

This is the strongest item on the list. No sign-up, no login, no credits, it opens in the browser and it genuinely produces 32 images in one pass. It runs on ads, which is where they make it back. Alongside the prompt there are art style presets, aspect ratio, and ready-made modifiers for camera, color and genre; you can save the ones you like to the page's own gallery.

Three things are missing. First, the wait: the 16 and 32 options stretch the time significantly and the queue is shared with everyone, so don't ask for 32 images at peak hour and expect them back in two minutes. Second, resolution: typical output runs 512 to 1024 pixels, which is small for a YouTube thumbnail but plenty for references and variations.

Third and most important: there's no fixed model behind it. The prompt goes to a third-party backend, usually a hosted Stable Diffusion derivative. The page's owner can swap that backend, tighten the filter, or take the page down entirely. There's no quality guarantee either. Nobody is promising that the style you like today will still be there in three months.

On Perchance the address matters, there are copycat pages

Because anyone on Perchance can publish their own generator page, there are dozens of image generators with nearly identical names and different filters and backends. Don't arrive via search, use the link above. Also, the commercial rights on what you save are unclear; if it's going into a thumbnail, an ad or a product you sell, treat the output as a reference rather than the final asset.

MovieFlow: the "faceless channel" line doesn't hold here

The tool works as advertised: you type an idea, it builds a multi-scene story, generates the characters and the transitions. What isn't advertised is the order of the screens. You don't go straight from prompt to video; there's a storyboard step in between, and that's where the actual fixing happens.

1
Write the idea
Scene order and tone go here, duration does not
2
Pick the settings
Language, duration, aspect ratio, style. The language option lives here
3
Fix the storyboard
This is the skipped step. You settle the scenes here, before generating
4
Generate and wait
A few minutes per clip, longer at peak hours

Two things for setting expectations right: generation is slow, a single good-looking scene can take several minutes, and it doesn't hand you a long piece as one file. For a longer story you'll need an editor outside the tool to assemble the parts. On the free side the duration options are short, hovering around eight seconds, one minute and two minutes.

The free plan isn't enough for a channel earning ad revenue

The use case in the original list is a faceless YouTube channel. On the free plan you can build and preview freely, but the file you download arrives watermarked, removing the watermark spends credits, and free output isn't cleared for commercial use. So you can't do the job the list suggests at the price the list quotes. The free plan is exactly enough for trying it out, not for publishing.

I'm not quoting a price

Plenty of sites list MovieFlow's plans and their numbers don't agree: different plan names and different credit amounts for the same product across different pages. I couldn't verify any of them at the source; they're all third parties. An unverified number doesn't go on this page. You'll see the price inside the app on your own account, so go look before you build the story.

MovieFlow

A multi-scene cartoon from one prompt. It builds the characters, scenes and transitions itself and puts a storyboard step in between. On the free plan you can build and preview; the file you download arrives watermarked.

Sample prompt

The story of a cat learning to skateboard.
Scenes in order: the park, the street, the backyard.
Each scene should follow from the last, with the cat getting slightly better every try.
Tone: bright colors, exaggerated expressions, wordless comedy.
Ending: the cat lands without falling for the first time and looks at the camera.

Don't write duration, language or aspect ratio into the prompt, pick them on the settings screen. The language option is there too. Spend the prompt on scene order and tone, because that's the part the model actually reads.

Open

Arena

Formerly LMArena. It runs the same prompt on two models at once and shows the outputs side by side. There are separate leaderboards for text, image, video and image editing, and on the video side you get free generations from frontier models while you vote.

Battle: two anonymous models compete, names revealed after the vote
Side by side: you pick the two models, votes don't feed the leaderboard
Direct chat: a single model, no voting

Sample prompt

A cinematic sunset over a futuristic city skyline.
Low angle, long lens, a sky running from orange into purple.
Last light reflecting off the glass facades, slow-moving air traffic in the distance.
The camera slowly tilts up.

Don't write the number of models into the prompt, model choice happens on screen. More importantly: what you type here can be disclosed to model providers and published in public research datasets. Keep client work, unpublished ideas and personal information out of it.

Open The leaderboards

Perchance

An image generator with no sign-up, no login and no credits. It produces up to 32 images at once and runs on ads. The fastest way to pull variations for thumbnails, social posts and inspiration boards.

Sample prompt

A futuristic robot mascot holding a coffee cup,
bold-outlined sticker style, flat single-color background,
soft side lighting, lightly metallic surface.

The count isn't written into the prompt, it's chosen from the image count dropdown under the box: 2, 4, 6, 8, 16, 32. The default is 4, so if you touch nothing you get 4 images, not 32. The 16 and 32 options noticeably lengthen the wait. Output runs 512 to 1024 pixels, which is small for a thumbnail.

Open

DOA: Yapay Zeka ve Otomasyon

Installing these tools on your own is one thing; actually building with them is another. The community has people using these daily and people building systems from scratch.

If you want a system that actually runs in your business, let's talk for 10 minutes; I'll look at what you're trying to build and tell you which path fits. Free, and not a sales pitch.

Book a 10-minute call

If you'd rather learn this alongside people doing the same work instead of on your own, the community is always open:

Join the community

This is an affiliate link.

← All resources