Àkọlé àwòrán

Àwọn Àkọlé

Fi àwòrán ojú àti àwòrán ìrànwọ́ pamọ́ - gba àwòrán ìrànwọ́ tí o bá sọ̀rọ̀ nípa ìṣàfarawé àwọn ojú, ìṣàfihàn ojú, àti àwọn ìṣàfihàn. Tí SadTalker (MIT) ṣe. Lórúkọ́ ìṣowó Òkè.

Ṣàfihàn

A kò ní àwọn ìrànwọ́ TTS nínú ìtàn rẹ̀. Yọ̀ọ̀kan rán wà láti fàyè gba àwọn rẹ̀! Fi Ojú Rẹ̀ pamọ́

Fi Àwòrán Pamọ́

Àwọn àyọkà 1,000 nínú ìṣísẹ̀ kan

1. Àwòrán Àwòrán àti Àwòrán Ìjáde

Gbé àwọn fáìlì rẹ̀ lọ́wọ́lọ́wọ́ síbẹ̀, tàbí Wá

JPG, PNG, or short MP4/WebM. Max 10MB. One clear, well-lit face works best.

2. Ìjánú àwòrán

Gbé àwọn fáìlì rẹ̀ lọ́wọ́lọ́wọ́ síbẹ̀, tàbí Wá

MP3, WAV, M4A, or FLAC. Max 10MB. Free: up to 30 sec. Pro: up to 5 min.

Àwọn ìkúndùǹ ìṣàfarawé àwọn ìṣàfarawé àwọn ìṣàfarawé

Ìwọ̀n Àwọn Àtòjọ-ẹ̀yàn

Àwọn

GFPGAN (tí o bá jù lọ́wọ́lọ́wọ́ lọ́wọ́lọ́wọ́)

Àwọn Àkọlé

SadTalker (CVPR 2023, Tencent ARC) ní módè́lì ìraǹrọ̀-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-ìgbà-

Code and weights are MIT-licensed end to end — no Llama, Gemma, or non-commercial backbone — so the videos you generate are safe for commercial use.

Àwọn Àlàyé fún Àwọn Àtòjọ-ẹ̀yàn

Lo àwòrán ìṣàfarawé ìṣàfarawé tí a bá mọ́ - ojù náà tí a rí, ẹnu tí a tí pa
Centered face, square or 4:5 aspect ratio works best
Àwọn àwòrán ìsàlẹ̀-ilà ìṣàfarawé (kò ní ìranṣẹ́) náà náà ìṣàfarawé àwọn àwòrán
Fún GFPGAN láti lo fún àwọn àkókò àìdájú - gbàdúró àkókò ìṣàfihàn láti fi àwọn àwọn ìròyìn hàn
Lo àtòjọ-ẹ̀yàn ìṣàfihàn tí a tí ìṣàfihàn àwọn àwòrán

Àwọn Àtòjọ-ètò Àwọn Àkọsílẹ̀

Ṣílẹ̀ ọ̀fẹ́, ṣíṣàfihàn tí o bà fẹ́ diẹ́ sii

Àìfẹ́

Ìgbà ìpamọ́ àwòrán 30-sekondì
Àwọn ààtò ìjánu-ìṣàmúlò-ètò
"Still" àwọn ààtò àìṣàfilọ́lẹ̀
Kò ní ìṣàfihàn ojù

Àwọn Àkọ́gbégbé

Àwọn

Ìgbà ìpamọ́ àwòrán 30-sekondì
Àwọn ààtò àwọn ìṣàmúlò-ètò "full" àti "still"
256 / 512 px ìjánu-ìṣàfihàn
Àwọn àwọn ìṣàfihàn àwọn oju GFPGAN

Ṣẹ̀dà

Àwọn

Ìgbà ìpamọ́ àwòrán 5-minún
Àwọn ìṣàmúlò-ètò GPU ìṣàfarawé
Ìyàn API (àfihàn àwọn ààyè-iṣẹ́)
Àwọn ìṣàfarawé àwọn ìṣàfarawé àwọn àgbéwọlé
Ìlò àìṣé (láìsì MIT)

_Ṣàfikún

Àwọn Àtòjọ-ẹ̀yàn

Upload a face photo and an audio clip, and the AI generates a video of that face speaking the audio with realistic lip movements, head pose, and blinks. Built on SadTalker (CVPR 2023), an MIT-licensed talking-head model that animates expression in addition to mouth shape.

Àwọn ìṣàfilọ́lẹ̀ ojù lè jẹ́ àwòrán JPG tàbí PNG (tí 10 MB lọ́wọ́lọ́wọ́) tàbí àwòrán àwọn ààyè-iṣẹ́ MP4/WebM tí a fi pamọ́ (a lò àwọn fèrèsé akọkọ). Àwòrán tí a fi pamọ́ lè jẹ́ MP3, WAV, M4A, tàbí FLAC tí 10 MB lọ́wọ́lọ́wọ́. A tí yipada àwòrán láti inú lọ́wọ́lọ́wọ́ sí 16 kHz.

Free accounts: up to 30 seconds per clip. Paying users: up to 5 minutes per request. Longer audio means longer render time and higher character cost.

Lip sync video uses 1,000 characters per second of generated video. A 30-second clip = 30,000 characters. The cost is billed up front from your character balance and refunded automatically if generation fails.

Ya — Àwọn ìṣàmúlò-ètò SadTalker àti àwọn ìṣàmúlò-ètò àwọn àwọn ìṣàmúlò-ètò ní ìṣẹ̀dá MIT tí a fi pamọ́ sí ìparí (kò si Llama, Gemma, tàbí àwọn ìrísí-lẹ́tà tí kò jẹ́ fún iṣowo). Àwọn àwọn àwòrán tí o ṣẹ̀dá ní ẹ̀yàn rẹ̀ láti lo fún iṣowo. Ò ní ìṣẹ̀dá láti ní àwọn ìṣẹ̀dá àwọn àwòrán ojú àti àwòrán àwòrán tí o fi pamọ́.

About 30 seconds for a 5-second clip on our A100 server, scaling roughly linearly with audio length. Enabling the GFPGAN face enhancer roughly doubles render time but produces sharper, higher-quality output.

Àwọn ààtò ìṣàmúlò-ètò ìparí iṣẹ́ (pípálẹ̀) àwọn ìṣàmúlò-ètò ìṣàmúlò-ètò ìṣàmúlò-ètò ìsàlẹ̀-ètò, àwọn ìṣàmúlò-ètò ìṣàmúlò-ètò, àti àwọn ìṣàmúlò-ètò nípa àwọn ojú, láti mú àwọn àwòrán ìsàlẹ̀-ètò ìsàlẹ̀-ètò tí a bá sọ̀rọ̀ pọ̀ jú lọ. Àwọn ààtò ìsàlẹ̀-ètò tí a tí wà nípa àwọn ààtò ìṣàmúlò-ètò ìsàlẹ̀-ètò náà láti pa ojú náà nípa àwọn ojú náà.

GFPGAN is a face restoration model that sharpens facial details after lip-sync rendering. It cleans up artifacts and makes 256-pixel output look closer to 512. It roughly doubles render time but is worth it for hero shots.

SadTalker renders at 256 px by default. Switch to 512 px size for sharper output (slower, higher VRAM) or enable the GFPGAN enhancer to upscale facial details. For best results, upload a high-quality, well-lit portrait photo.

Yes. Upload an MP4 or WebM as the face input and we will use the first frame as the driving identity. For full video re-dubbing (per-frame mouth replacement), see the upcoming Dubbing Studio video pipeline.

Ya. POST àgbèwọlé àwọn ààyè-iṣẹ́ àwọn àwọn ààyè-iṣẹ́ àwọn àwọn ààyè-iṣẹ́ àwọn àwọn ààyè-iṣẹ́ àwọn àwọn ààyè-iṣẹ́ àwọn àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-iṣẹ́ àwọn ààyè-

SadTalker uses face-alignment to detect and crop the most prominent face. For best results, upload a portrait with one person centered, eyes visible, and minimal occlusion. Group photos may produce unpredictable results.

5.0/5 (1)

Tí o bá fẹ́ bẹrẹ?

Ṣẹ̀dá ọ̀fẹ́ ki o si gba 50 kọ́rẹ́dì. A kò nilò kaadí ẹ̀yàn.

Ṣàfihàn Wó Àwọn Ìtàn

Àwọn Àkọlé

Fi Àwòrán Pamọ́

Àwọn Àwòrán Àwọn Àkọ́kọ́ Rẹ̀

Àwọn Àkọlé

Àwọn Àlàyé fún Àwọn Àtòjọ-ẹ̀yàn

Àwọn Àtòjọ-ètò Àwọn Àkọsílẹ̀

Àwọn Àtòjọ-ẹ̀yàn

Àwọn Ìṣàmúlò-ètò Àìṣàmúlò-ètò AI

Àwọn ìrísí-lẹ́tà wo ní a fọwọ́sì?

Ìgbà wo ní àwòrán náà lè jẹ́?

Igba wo ni o n lọ?

Ń lè ló àwọn àwòrán nípa ìṣowo?

Igba wo ni a fi gba àwọn ìṣàfarawé?

Kini ìfaràǹwọ́ laarin àwọn ààtò àwọn ààtò "full" ati "still"?

Kini aago GFPGAN?

Kini idi ti àwọn ààtòjútó mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn mìíràn

Ń lè fi àwòrán pamọ́ sí àwòrán tuntun?

Àwọn Ìṣàmúlò-ètò

Njẹ́ tí àwòrán ojú mì bá ní àwọn eniyan mìíràn nínú rẹ̀?

Tí o bá fẹ́ bẹrẹ?