Live Portrait (performance capture)
Free.ai
·
video
·
~500 Ama-token ngalinye video clip
·
3.2
kusuka 4 abasebenzisi balolu hlu
Live Portrait (performance capture) yi an imodeli yokukhiqiza ividiyo. Ihamba ngezindlela zangaphandle — ~500 i-token ~5s ngasinye (50% yokuphawula ngaphezulu kwezindleko eziphezulu).
Sebenzisa nge-API
curl -X POST https://api.free.ai/v1/video/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"premium/live-portrait","prompt":"your prompt here"}'Imibuzo ebuzwa kaningi
Live Portrait (performance capture) yisinye sezimodeli zevidiyo ze-AI ezitholakala ku-Free.ai. Sebenzisa ifomu kulekhasi ukuletha i-clip; ngaphansi kwe-fold uzothola iziphakamiso zesampula, ukuthengiselana, kanye nezinketho.
Ku-Free.ai, ukukhishwa kwevidiyo kufinyelela kumasekondi angama-6 ngeclip ngayinye kuzo zonke imodeli yevidiyo. Ukudlulisa, sebenzisa inkinobho yokudlulisa kuhele /ividiyo/ ithuluzi — ithatha ifreyimu yokugqibela futhi iqhubeka khona.
16:9 landscape (YouTube, desktop), 9:16 portrait (TikTok, Reels, Shorts), kanye ne-1:1 square (Instagram feed). Khetha isilinganiso kwifomu ngaphambi kokuba udale ukuze imodeli ibeke isithombe sohlobo.
720p HD ngokuzenzakalela. Ezinye imodeli eziphezulu zixhasa kuze kube yi-1080p — hlola ikhadi lokuqhathaniswa. Ungakhuphula noma yikuphi okukhishwayo ku /video/upscaler/ ngemuva kweqiniso.
Kunamandla ku-natural motion (abantu behamba, izimoto zihamba, amanzi, isimo sezulu). Kuncane ku-intricate hand interaction ne-fine text. Sebenzisa i-camera-direction hints (slow pan, dolly forward) ukuhlanza i-natural motion.
Free.ai inikeza ukusetshenziswa kokuthengiswa kwevidiyo ekhiqizwe kuwe kungakhathaliseki imodeli. Imodeli ngokwayo ingabe yi-Apache 2.0, MIT, noma ilayisense elikhethekile lomthengisi — bona i-imodeli yokubhekisa ikhonkco ngezansi.
Live Portrait (performance capture) iyimodeli yevidiyo esezingeni eliphakeme. Izindleko zilinganiselwa ngokuqhubekayo — ngokuvamile ~5,000 amathokheni ngomzuzu. $1 ithenga amathokheni angama-750,000, ngakho-ke iphakethe lama-tokens angama-$5 lifaka ama-clip amaningi. Ababhalisile abamahhala bathola amathokheni angama-10,000 ukuzama.
Yebo — sebenzisa /batch/ kuze kube yi-100. I-clip ngayinye ifakwe eqenjini futhi imiphumela iwela ku /account/?tab=history uma ilungile.
Yebo — thumela umlayezo wakho ku /v1/video/generate/ ngemodeli "Live Portrait (performance capture)" (noma sebenzisa i-slug kulekhasi). Bona /api/ ukubheka indawo yokuqeda egcwele kanye nokuqinisekisa, amakhodi wephutha, nama-SDK snippets.
Sigcina izicelo zakho ngaphandle kwefasitela lokuhlukaniswa (24h anon, 7d ukungena). Siqeqesha ngemingeniso yakho. Amamodeli abekwe ngokwezifiso asebenza ku-Free.ai-owned GPUs.
Yebo - imisebenzi engaphumelelanga zibuyiselwa ngokuzenzakalela kumthombo kuqala ukukhokha (i-pool yansuku zonke noma i-token ekhokhelwa) ngomhla owodwa. I-imeyili contact@free.ai uma ukubuyiselwa imali akubonisi.
Isibonelo sokubonisa 30-120 imizuzwana yevidiyo yemizuzu emi-4; amamodeli aphezulu avame ukuqedwa emizuzwani engu-15-45. Sebenzisa inkinobho yokuvala-i-tab-ephephile ye-queue ku /video/ — imiphumela yakho iwela ku /account/?tab=history uma yenziwe, nge-imeyili ekhethiwe.