Question 1

Gịnị bụ AI Depth ka inyogo?

Accepted Answer

Wepụta inyogo nlebara anya. AI na-ejikwa map n'ime (n'ihu ngwe ngwe, ihe dị nso na ihe dị anya site na kamera) nakwa iji 3D nhazi dịka nkwụsị mgbe ị na-egosipụta inyogo ọhụrụ n'ụdị ọbụla mọọbụ ihenhọrọ ịkọwapụtara.

Question 2

Kedụ mgbe m ga-eji ụkpụrụkpụ versus. canny versus. pose?

Accepted Answer

Depth na-echekwa SCENE GEOMETRY (ụlọ, kamera angle, n'ebe dị nso). Canny na-echekwa OUTLINE. Pose na-echekwa BODY POSITION. Jiri depth mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụla mgbe ọbụ

Question 3

Olee otú map ụkpụrụkpụ ahụ ga-esi pụta?

Accepted Answer

MiDaS (Intel ISL) - a monocular depth estimator. O na-ele 2D inyogo na-atụ anya per-pixel depth, enweghị 3D kamera ma ọ bụ LiDAR chọrọ. Depth map na-aghọ SDXL Depth ControlNet input.

Question 4

Kedu ụdị inyogo nlekọta ka mma?

Accepted Answer

Ihe ọbụla na-egosi n'ihu/okpuru n'otu n'otu. N'ime ụlọ, ihenhọrọ, n'èzí, n'ime ụlọ, ọbụla na-arụ ọrụ. Flat compositions (eg, poster graphic) na-enye maps nke n'ime ụlọ na-enweghị mmasị; nlekọta n'ime ụlọ bụ nke bara uru mgbe isi mmalite nwere 3D arụpụtaghị.

Question 5

Ọ bụ n'efu?

Accepted Answer

Ee. ~1,200 tokens kwa render (1,000 base + 20% ControlNet surcharge). Ndị ọrụ banye-na-enweta 5K n'efu kwa ụbọchị - banyere 4 depth-controlled renders kwa ụbọchị na free tier.

Question 6

M ga-eji ya mee ka ebe ahụ pụta?

Accepted Answer

N'ụzọ na-enweghị isi - map nke n'okpuru na-echekwa jiometrị, nkuzi na-ejikwa ọkụ ("n'ụbọchị", "n'okpuru ọkụ studio dị njọ", "site na ọkụ candlelight"). Ọnụọgụgụ na-echekwa nhazi ebe, na-atụgharị ọkụ cues. Ọ dị mma maka "ụlọ otu, oge dị iche iche nke ụbọchị" mmetụta.

Question 7

Olee otú ọ dị iche na img2img?

Accepted Answer

img2img na-eweregharị-eweregharị n'elu input n'ụzọ ziri ezi - agba, etiti, na jiometrị. Ogo-na-ịhoputa-ịhoputa agba na etiti, na-echekwa naanị jiometrị. Ọrụ dị ka nke ahụ, ihenhọrọ ọhụrụ. img2img na-adabere n'ihe; ogo na-enye gị ohere ịgbanwe isiokwu.

Question 8

Enwere m ike ịtọgharịa ka ọ dị ka ọbụla n'ime mmiri ahụ?

Accepted Answer

Ya - ControlNet ike slider. Dìfọ́ọ̀ltụ̀ 0.7. 1.0 = n'obere n'obere (ọ̀pụ̀pụ̀ dị ka n'ime-mapụ́gharịrị-n'ime-n'ime); 0.4 = gbasaa (ọ̀dịgasị iche iche nke n'ime). Túnù n'ime ngosi.

Question 9

Kedu ihe bụ ụhara pụtapụta a na-akwado?

Accepted Answer

512×512 dìfọ́ọ̀ltụ̀. Ruo 1024×1024 na kpụrụkpụmkpụ SDXL aspect ratios. Ụhara ndị dị ukwuu na-eji VRAM + tokens ndị ọzọ.

Question 10

Ọ bụrụ na ọbụla, ọ ga-egosi mapụ ụkpụrụkpụ?

Accepted Answer

Ụnyaahụ, ọbụla inyogo nke e megharịrị na-ebuli. N'oge a, a na-ejikwa map nke n'obosara na GPU + na-ewepụ. Ọ bụrụ na ịchọrọ ịkọwapụta n'obosara nke n'obosara dịka mbido dị iche iche, anyị nwere ike ịkpọgharịa na - faịlụ n'ihe n'ime /contact/.

Question 11

Ọ bụrụ na a chekwara ndesịta ozi ahụ?

Accepted Answer

E mepụtara ya n'oge na-adịghị anya, e weghachiri ya n'ime, wee wepu ya. Ọ bụ naanị profaịlụ gị + ngwụcha render ka echekwara na /akaụntụ/?tab=ihe ọmụma. A na-ejighị ya maka nkuzi. /privacy/ maka usoroiheomume zuru ezu.

Question 12

Ọ bụ API?

Accepted Answer

Ee - POST multipart ka /v1/image/generate/ na model=sdxl, nkuzi, control_image (faịlụ), control_type=depth. Onye na-eweta ikike, 10K free tokens/ọnwa. Curl na /api/.

Depth → Image (ControlNet)

Ihenhọrọ ahụ

Depth → Image (ControlNet)

Ihenhọrọ ahụ

30,000 Free Tokens, kwa ụbọchị

Nwanyị: 30K free tokens/abalị!

Wanọbụla?