SadTalker
Free.ai (self-hosted)
·
video
·
~30000 Ii-token video clip
·
3.2
I- 78 Abathathi nxaxheba
SadTalker yi an imodeli yokudala ividiyo eyenziwe ngu OpenTalker. Enamandla kakhulu kwi Avatar explainers, voice-over to video, talking-head ads, dubbed content.. I-self-hosted kwi-Free.ai GPUs — iqhuba ngokukhululekileyo kwi-token pool yakho yosuku (30,000 tokens kwi ~5s iqhosha ngalinye). Ikhutshwe phantsi kwe Non-commercial (BFM) — ukusetyenziswa korhwebo kuvunyelwe kwi Free.ai.
Sebenzisa nge API
I-REST API ehambelana ne-OpenAI. Yenza iqhosha kwaye unxulumane nale modeli kwimizuzu.
curl -X POST https://api.free.ai/v1/video/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"sadtalker","prompt":"your prompt here"}'Iimodeli ezifanayo
Imibuzo ebuzwa rhoqo
SadTalker yenye yeemodeli zevidiyo ze-AI ezifumanekayo kwi-Free.ai. Sebenzisa ifom kule phepha ukudala iclip; ngaphantsi kwe-fold uya kufumana iimfuno zesampuli, ukuthengiswa, kunye neendlela ezikhethwayo.
Kwi Free.ai, ukwenziwa kwevidiyo kuphezulu kuphuma kwimizuzu emi-6 ngeclip nganye kuzo zonke iimodyuli zevidiyo. Ukuwunyusa, sebenzisa iqhosha lokunyusa kwisixhobo esipheleleyo /ividiyo/ — ithatha isakhelo sokugqibela kwaye iqhubeka ukusuka apho.
16:9 ubuhle belizwe (iYouTube, idesktop), 9:16 ubume bendawo (iTikTok, iReels, iShorts), kunye ne-1:1 isikwere (i-Instagram feed). Khetha ubungakanani befom phambi kokuba uvelise ukuze imodeli ifake i-frames ye-shot ye-shape.
720p HD ngokumiselweyo. Ezinye iimodyuli eziphezulu zixhasa ukuya kwi-1080p — khangela ikhadi lokuthelekiswa. Unganyusa nayiphi na i-output kwi /video/upscaler/ emva kwenyameko.
Inamandla kwintshukumo eqhelekileyo (abantu behamba, iimashini ziqhuba, amanzi, iimeko zemozulu). Inamandla kwiintshukumo zesandla ezinzima kunye nombhalo olungileyo. Sebenzisa iincam zekhamera (i-slow pan, i-dolly phambili) kwintshukumo ecocekileyo.
Free.ai inika ukusetyenziswa korhwebo kwevidiyo eziveliswe kuwe ngaphandle kokuxhomekeka kwimodeli. Imodeli ngokwayo ingaba yi-Apache 2.0, MIT, okanye ilayisenisi ekhethekileyo yomthengisi - bona ikhonkco lokubhekisa kwimodeli ngezantsi.
SadTalker runs on our own GPUs at 30,000 tokens per clip, flat: a render costs what it costs because of model load and diffusion steps, not clip length, and 2,500 per second only takes over past 12 seconds. That is the entire 30,000-token daily free pool, so a free account gets one clip every two days. Paid token packs start at $5 → 200,000 tokens and remove the cap.
Ewe - sebenzisa /batch/ ukuya kuthi ga kwi-100 ye-prompts. Iclip nganye ifolokhwe kwaye i-result iwela kwi-/account/?tab=history xa ilungile.
Ewe — UTHENGA umyalezo wakho okhawulezayo ku /v1/video/generate/ ngemodeli="SadTalker" (okanye sebenzisa i-slug kule phepha). Bona /api/ ukubhekisa kwinto egqithileyo kunye nogunyaziso, ikhowudi yemposiso, kunye ne-SDK snippets.
Asikho kukugcina izilumkiso zakho ngaphandle kwefestile yokwabelana (24h anon, 7d ubhaliso-ngaphakathi). Asiqeqeshi kwingeniso zakho. Iimodeli ezimkelweyo zisebenza kwi-Free.ai-ekhoyo kwi-GPUs.
Ewe — imisebenzi engaphumelelanga ibuyiselwa ngokuzenzekelayo kumboneleli wemali osemthethweni (i-pool yosuku ngalunye okanye i-token ehlawulwe) ngomhla owodwa. I-imeyili contact@free.ai ukuba ubuyiselo alukho.
Ukubonisa okuqhelekileyo 30-120 imizuzwana ye clip yemizuzu emi-4; iimodyuli eziphezulu zihlala zigqiba kwimizuzu emi-15-45. Sebenzisa iqhosha lokuvala-i-tab-ekhuselekileyo yofolo kwi /vidiyo/ — isiphumo sakho siza ku /i-akhawunti/?tab=imbali xa kwenziwe, nge-imeyili ekhethiweyo.