{"id":25427,"date":"2024-12-21T21:04:30","date_gmt":"2024-12-21T13:04:30","guid":{"rendered":"https:\/\/www.1ai.net\/?p=25427"},"modified":"2024-12-21T21:21:07","modified_gmt":"2024-12-21T13:21:07","slug":"ai%e8%a7%86%e9%a2%91%e5%b7%a5%e5%85%b7%e5%88%b0%e5%ba%95%e5%93%aa%e4%b8%aa%e5%a5%bd%e7%94%a8%ef%bc%9f%e5%93%aa%e4%b8%aa%e5%88%92%e7%ae%97%ef%bc%9f6%e6%ac%be%e5%9b%bd%e5%86%85%e5%a4%96ai%e7%94%9f","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/25427.html","title":{"rendered":"AI video tools in the end which is good to use? Which one is cost-effective?6 domestic and international AI generation video tools comparison test"},"content":{"rendered":"

There are various (closed source)AI Video Tools<\/a>, dazzling. Which one works? Which one is cost-effective?<\/p>\n

Three domestic tools were selected today:Keling<\/a>,Dream<\/a>,whelks<\/a>, and three foreign tools: sora, runway, and luma, for a side-by-side comparison.<\/p>\n

\"437341c2j00sougz6000ld000dm00akm\"<\/p>\n

All tools use the latest models, including version 1.6, which has just been released by Koring:<\/p>\n

\"42fa242bj00sougz6000hd000j4004sm\"<\/p>\n

Test Method: Using 720p\/5s video supported by all tools, 4 samples were generated for each test, each tool, and the best 1 of them was selected for comparison.<\/p>\n

I. Vincennes Video<\/strong><\/p>\n

Overall test video<\/p>\n

Case 1:<\/p>\n

\"4058722cg00sougz803xvd000go00gom\"<\/p>\n

The sora-generated cat repeatedly jumps across the room, only partially completing the cue word.<\/p>\n

\"c0de47bdj00sougz6000rd000in007ym\"<\/p>\n

Case 2:<\/p>\n

\"47e668f7g00sougz803xed000go00gom\"<\/p>\n

All the tools failed to properly represent the dragon breathing fire from its mouth.<\/p>\n

\"a8cb9392j00sougz60011d000ky008um\"<\/p>\n

Case 3:<\/p>\n

\"bfbee3dfg00sougz803qod000go00gom\"<\/p>\n

Half the tools don't understand what a \"first person shooter\" is.<\/p>\n

\"0dff64f1j00sougz6000vd000js008um\"<\/p>\n

Case 4:<\/p>\n

\"9f6b6127g00sougz703odd000eg00egm\"<\/p>\n

None of them are perfect if you look closely, such as men in women's swimsuits or chipped faces. All things considered, luma fared the worst.<\/p>\n

\"8df65273j00sougz6000yd000j2009im\"<\/p>\n

Case 5:<\/p>\n

\"469b18eeg00sougz8043od000go00gom\"<\/p>\n

All of the tools performed well, with some missing details.<\/p>\n

\"4cd3d550j00sougz6001ed000pu00cfm\"<\/p>\n

Case 6:<\/p>\n

\"1353f038g00sougz804bcd000go00gom\"<\/p>\n

Sora failed to reflect the \"dirt road\" and Luma was again the worst performer.<\/p>\n

\"fbcca4e3j00sougz6001id000or00erm\"<\/p>\n

Case 7:<\/p>\n

\"b1d51835g00sougz702y0d000fa00fam\"<\/p>\n

All the tools performed well, but the Spirit did not understand the \"flame composition\" properly.<\/p>\n

\"588830b3j00sougz60013d000u000anm\"<\/p>\n

Case 8:<\/p>\n

\"a17ad250g00sougz703fvd000fa00fam\"<\/p>\n

Famous movie and TV IPs are hard to get satisfactory results through text-to-life videos, so it is recommended to try graphic-to-life videos.<\/p>\n

\"f1ea0c26j00sougz6001id000ti00alm\"<\/p>\n

Case 9:<\/p>\n

\"06993187g00sougz7029vd000eg00egm\"<\/p>\n

This set of cue words is quite difficult, with Korin and Sora's results barely usable, and RUNWAY and LUMA shortchanging the characters' movements significantly.<\/p>\n

\"bf5cdc61j00sougz60017d000u0009lm\"<\/p>\n

Case 10:<\/p>\n

\"a5519a21g00sougz602qsd000go00gom\"<\/p>\n

Celebrity stars are difficult to generate, and if the characters don't match, the whole piece is meaningless. All the tools failed except for the conch.<\/p>\n

\"942c8a2fj00sougz6001bd000u000bdm\"<\/p>\n

II. Toussaint video<\/strong><\/p>\n

Overall test video<\/p>\n

Case 1:<\/p>\n

\"ea1549bej00sougz60014d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"0557f4bag00sougz6016ld000fa00fam\"<\/p>\n

Taken together, korin and imouto perform the best. sora's characters don't move but run mirrors haphazardly, and cue word compliance is poor.<\/p>\n

\"fd223096j00sougz60013d000t3008fm\"<\/p>\n

Case 2:<\/p>\n

\"3f1d503aj00souhas0022d000u000ggm\"<\/p>\n

Test results:<\/p>\n

\"ffc34f4cg00souhas029qd000fa00fam\"<\/p>\n

Korin exceeded expectations in this test. sora was good, but purely \"random\". Half of the tools do not reflect the launch and explosion effects.<\/p>\n

\"5e1cbeb2j00sougz60018d000nr00aom\"<\/p>\n

Case 3:<\/p>\n

\"bb2f443aj00souha7001cd000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"a9acb288g00souha802bmd000fa00fam\"<\/p>\n

Conch was good (although the pedestrians still had chipped limbs), Korin and Ieomu were okay, and RUNWAY's artwork felt a bit poor.<\/p>\n

\"cb1cde3cj00sougz60018d000u000ahm\"<\/p>\n

Case 4:<\/p>\n

\"8b5704b0j00souh9q0026d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"5d102160g00souh9q02jqd000dw00dwm\"<\/p>\n

Conch and luma morphed the guns into barricades, and fortunately most of the tools played well.<\/p>\n

\"61602de2j00sougz60013d000nc00aom\"<\/p>\n

Case 5:<\/p>\n

\"5356dbabj00souh960025d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"dab1b92dg00souh9602did000fa00fam\"<\/p>\n

Character movement is a major test for video tools, and Korin and Conch double over.<\/p>\n

\"acfb81e4j00sougz60014d000n900alm\"<\/p>\n

Case 6:<\/p>\n

\"5111165ej00souh8n0025d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"79db6952g00souh8o01p5d000fa00fam\"<\/p>\n

Korin has improved its dynamic graphics after upgrading to 1.6, and this time it reigns supreme with Conch.<\/p>\n

\"af18e6cdj00sougz60011d000pf008im\"<\/p>\n

Case 7:<\/p>\n

\"9e2a3a9bj00souh81003ad000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"25edc723g00souh810424d000fa00fam\"<\/p>\n

If you look closely, scenes like this have more or less faces or bodies falling apart, but luma looks weaker in comparison.<\/p>\n

\"233aeb78j00sougz6000xd000ou008im\"<\/p>\n

Case 8:<\/p>\n

\"f8f01a8bj00souh7c0043d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"27eb114dg00souh7c031ed000eg00egm\"<\/p>\n

A relatively simple test that everyone completed with no problem. sora seems to have a preference for adding substantial movement on her own, and it's not out of place here.<\/p>\n

\"1d21370bj00sougz6000vd000nc008im\"<\/p>\n

Case 9:<\/p>\n

\"924a5d59j00souh6e0021d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"e48ed4fag00souh6e01wzd000dc00dcm\"<\/p>\n

Representing the landscape is relatively simple and most tools work smoothly.<\/p>\n

\"ec0a25c6j00sougz7000yd000nu008im\"<\/p>\n

Case 10:<\/p>\n

\"17c12e8ej00souh5o0032d000u000gtm\"<\/p>\n

Test results:<\/p>\n

\"0aee4835g00souh5o02y3d000eg00egm\"<\/p>\n

Once again, sora appears to switch scenes abruptly. On the contrary, several domestic tools performed well.<\/p>\n

\"0f8f3722j00sougz7000yd000mr008im\"<\/p>\n

III. Comparison of the performance\/functionality of the tools<\/strong><\/p>\n

Performance: \"High Definition + Longer + More Concurrency\" is good, but it will cost more.<\/p>\n

\"fc7548d8j00sougz7001vd000pu00dwm\"<\/p>\n

Efficiency: korin and conch are the slowest, suggesting more servers.<\/p>\n

\"b2dae09ej00sougz7000nd000k7005km\"<\/p>\n

Vetting: the stricter the vetting, the greater the constraints on creativity, want to be free? Please use open source tools.<\/p>\n

\"b13452d7j00sougz7000td000lh007bm\"<\/p>\n

Functions: sora, korin, runway are all very rich in functions, but more functions do not necessarily equal to good use.<\/p>\n

\"e813f270j00sougz7001id000ml00exm\"<\/p>\n

Price: Conch's premium membership allows unlimited generation, awesome! All three foreign tools are expensive by comparison. (It cost me a lot of rice to complete this test)<\/p>\n

\"\"<\/p>\n

Summarize:<\/strong><\/p>\n

1. On balance, domestic tools win!The effects of Korin and Yumemon are very stable.<\/strong>, is preferred.Conch Dynamics is doing well<\/strong>.<\/p>\n

2.Sora is the most expensive and less effective<\/strong>However, considering that ChatGPT Pro can be activated at the same time, it can be used as a \"freebie\" for the latter.<\/p>\n

3. runway is a trailblazer and has quite a few exclusive features (like expression capture), butToo few runway points.<\/strong>, don't use the premium member's casual mode, it runs out quickly.<\/p>\n

4.luma is less effective and not recommended at this time<\/strong>.<\/p>\n

But AI modeling is dynamic, and even tools that lag behind today may turn over later, so stay tuned!<\/p>\n

 <\/p>\n

The article covers the URL:<\/p>\n

Sora:
\nhttps:\/\/www.1ai.net\/25487.html<\/a><\/p>\n

Can be spiritual:
\nhttps:\/\/www.1ai.net\/12558.html<\/a><\/p>\n

I.e., dreams:
\nhttps:\/\/www.1ai.net\/10005.html<\/a><\/p>\n

Conch:
\nhttps:\/\/www.1ai.net\/21428.html<\/a><\/p>\n

Runway:
\nhttps:\/\/www.1ai.net\/78.html<\/a><\/p>\n

luma:
\nhttps:\/\/www.1ai.net\/13078.html<\/a><\/p>","protected":false},"excerpt":{"rendered":"

There is a bewildering variety of (closed source) AI video tools on the market now. Which one is good to use? Which is cost-effective? Today selected three domestic tools: can spirit, namely dream, conch, as well as three foreign tools: sora, runway, luma, for side-by-side comparison. All tools use the latest model, including can Ling just released version 1.6: test method: using all tools are supported by the 720p\/5s video, each test, each tool are generated 4 samples, select the best 1 to compare. First, Vincennes video Overall test video Case 1: sora generated by the cat repeatedly horizontal jump, can only be counted as part of the completion of the cue word. Case 2: All tools failed to correctly represent the dragon breathing fire from its mouth. Case 3: Half of the tools didn't understand what the<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[144],"tags":[170,3598,195,1249,2531,2957,4870],"collection":[],"class_list":["post-25427","post","type-post","status-publish","format-standard","hentry","category-baike","tag-ai","tag-luma","tag-runway","tag-sora","tag-2531","tag-2957","tag-4870"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/25427","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=25427"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/25427\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=25427"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=25427"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=25427"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=25427"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}