Trao tao Giong Noi AI cho Kios Thanh toan Tu dong Retail
AI giong noi tu dong thanh toan bay gio la gương mat thính giác cua cua hang ban le hien dai. Moi lan mot khach hang nghe “vui long dat hang hoa vao khu vuc dong goi” tai mot lajur tu dong thanh toan Walmart, Kroger hoac Carrefour, giong noi do duoc tao ra boi mot he thong text-to-speech — va ngay cang nhieu, he thong do la mot tao giong noi AI thay vi mot ban ghi phong thu tu mot dien vien giong noi duoc thue. Huong dan nay giai thich cach cac nha ban le cau hinh giong noi kios tu dong thanh toan tren phan cung NCR Voyix va Diebold Nixdorf, nhung gi ma phuong phap tu do WCAG 2.1 tren thuc te yeu cau cho am thanh kios, cach cac thu vien yeu cau da ngon ngu duoc cau truc, va cach tao ra mot nhan vat giong noi thuong hieu phu hop hoat dong o 2.000 lajur trong mot chuoi.
TL;DR
- AI giong noi tu dong thanh toan dieu khien cac yeu cau audio tren cac kios tai Walmart, Kroger, Carrefour va hau het cac chuoi lon — “vui long dat hang hoa vao khu vuc dong goi” la vi du duoc nhan ra nhieu nhat.
- NCR Voyix va Diebold Nixdorf la cac nha san xuat OEM tru thong; ca hai deu su dung cac thu vien yeu cau WAV duoc load vao dieu khien may bay.
- WCAG 2.1 yeu cau moi yeu cau hinh anh phai co mot dieu chinh audio, co the hieu duoc o nhung am luong kios, voi dieu khien ng[dung tai audio.
- Cac kios da ngon ngu (Tieng Anh + Tay Ban Nha tai Walmart, Phap + Arab tai Carrefour) can cac thu vien yeu cau rieng biet theo ngon ngu tu cung mot pro-fil giong noi.
- Cac tao giong noi AI thay the cac phien ban tung phuong tung phien tro voi tao dot tu mot ky ban — quan trong o quy mo chuoi o do mot cap nhat yeu cau duy nhat cam trac hang ngan may.
- VoxBooster xu ly kloning giong noi va san xuat dot WAV cho cac quy trinh lam viec am thanh ban le tren Windows.
Nhung gi ma AI Giong noi Tu dong Thanh toan Thuc su La
AI giong noi kios ban le dua toi moi co cau no phat sinh cac yeu cau audio huong dan khach hang thong qua mot giao dich self-scan checkout. Cum tu “AI giong noi tu dong thanh toan” bao gom toan bo ngan xep: nhan vat giong noi chinh (da, da bieu, dang ki giong), thu vien yeu cau (moi dong ky ban co the ma he thong co the phat), dinh dang tep audio (tieu thuc WAV dieu khien chap nhan), va logic tao nhan lan yeu cau nao duoc phat khi.
Day duoi duc de tien hanh yeu cau dien roi o mot may bay tu dong thanh toan co tro nhu sau:
- “Chao mung. Vui long quet san pham dau tien cua ban.”
- “Vui long dat hang hoa vao khu vuc dong goi.”
- “Hang hoa khong mong doi trong khu vuc dong goi.” (la phat hien khong khop)
- “Co the ban co phieu giam gia hoac the loyalitas khong?”
- “Vui long chon phuong thuc thanh toan cua ban.”
- “Vui long chen the cua ban.” / “Vui long chum the cua ban.”
- “Vui long lay ra the cua ban.”
- “Giao dich duoc phe duyet. Vui long lay tham bien nhan va hang hoa cua ban.”
Moi dong la mot tep WAV rieng le trong thu vien yeu cau cua may bay. Mot thu vien hoan toan — bao gom tat ca cac trang thai loi, xac minh tuoi, tim kiem san pham, canh bao su khac biet can bang, cac yeu cau ghi de nhat cua co van, va cac thong diep dong cua — tro len den 80-150 clip rieng le tren moi ngon ngu tren moi loai lajur.
Nhan len toan bo chuoi ban le 500 cua hang voi 4 lajur tren cua hang va 2 ngon ngu, va ban co tro len den 1,2 trieu tep audio rieng le de tao ra, quan ly va cap nhat. Day la ly do tai sao tao dot AI thay the ghi am phong thu cho am thanh ban le doanh nghiep: khi mot quy dinh moi yeu cau mot ky ban xac minh tuoi cap nhat, mot he thong AI tao lai cac clip bi anh huong trong mot gio. Mot phien hop phong thu co chi phi hang ngay va hang ngan dollar.
Tieng noi Dung sau “Vui long Dat Hang hoa Vao Khu vuc Dong goi”
Yeu cau giong noi tu dong thanh toan duoc nhan ra nhieu nhat trong the gioi ban le tieng Anh la “vui long dat hang hoa vao khu vuc dong goi.” Trong hau het cac nam 2000 va 2010, giong noi do la mot ban ghi con nguoi — thong thuong la mot dien vien giong noi chuyen nghiep duoc thue theo co so hoac boi mot OEM phan cung (NCR hoac Diebold Nixdorf) hoac boi cac chuoi ban le lon de ghi am giong noi thuong hieu cua rieng ho.
Giong noi tu dong thanh toan Walmart US, vi du, da tro nen quen thuoc du de cum “hang hoa khong mong doi trong khu vuc dong goi” vao nhung khoa meme van hoa — mot dau hieu cua bao nhieu khach hang gap yeu cau nay, va suc manh cua su nhan dang thuong hieu am thanh.
Nhieu yeu to da day ma suy tu giong noi tho duoi sang giong noi tao ra AI:
Tan suat cap nhat. Cac he thong ban hang theo chieu doc cap nhat ky ban thuong xuyen — cac phuong thuc thanh toan moi, tao lai thuong hieu chuong trinh loyalitas, ngon ngu quy dinh cho cac mua ne ruou hoac thuoc, cac thong diep theo meua. Moi su thay doi ky ban truoc day yeu cau dat phong thu. Tao AI giam dieu nay xuong phut.
Quy mo toan cau. Cac nha ban le quoc te nhu Carrefour hoat dong tren hang chuc quoc gia va hang chuc ngon ngu. Thue tai nang giong noi ban dia chi cho moi ngon ngu tren moi thi truong, duy tri tinh nhat quan tren cac phien hop va quan ly cac hop dong tai nang o quy mo do phuc tap ve mac do hanh chinh. Tao giong noi AI xu ly tung ngon ngu tu mot ho so giong noi duoc xac dinh.
Tinh nhat quan cua thuong hieu. Mot nha ban le triển khai tu dong thanh toan tren 2.000 cua hang trong nam nam, su dung cac phien hop ghi am khac nhau khi chuoi phat trien, se ket thuc voi cac giong noi phat ra khong nhat quan tren cac co so — nhung cai tho hon, nhung cai giong co-bot nhieu hon, nhung cai co cac da khac nhau. Tao giong noi AI tu mot ho so duoc xac dinh tao dau ra giong het nhau tai may bay 1 va may bay 4.000.
Chi phi tren moi yeu cau. O cac ty le phong thu, mot thu vien yeu cau 120 clip trong hai ngon ngu chi phi hang ngan dollar. Tao AI giam chi phi bien dong cua cac yeu cau moi gan bang khong sau khi ho so giong noi duoc thiet lap.
Tu dong Thanh toan NCR Voyix: Kien truc Phan cung va Am thanh
NCR Voyix (truoc day la Cong ty NCR, tao lai ten hieu 2024) tao ra cac dong san pham FastLane, SelfServ 90 va EASY CHECKOUT ma ban tim thay tai Walmart, Kroger, Home Depot va hau het cac chuoi ban hang lon nhat o My. Hieu biet cach cac he thong nay xu ly am thanh la dieu quan trong doi voi bat ki ai tao giong noi kios tuy chinh.
Cac don vi tu dong thanh toan NCR FastLane va SelfServ chay Windows (thong thuong la Windows 10 IoT Enterprise tren phan cung tao ra hien tai) hoac mot OS tren co so Linux tren cac don vi cu hon. Am thanh duoc xu ly boi phan mem ung dung POS — Emerald POS NCR hoac nen tang SCOT (Giai phap Tu dong Thanh toan) — tao ra cac tep WAV tu mot thu muc thu vien yeu cau dia phuong tren may bay.
Tieu thuc audio cho cac he thong NCR:
| Dong NCR | Ty le Mau | Sau khi Bit | Kenh | Dinh dang |
|---|---|---|---|---|
| FastLane (tao ra hien tai) | 44,1 kHz | 16-bit | Mono | WAV PCM |
| SelfServ 90 | 22,05 kHz hoac 44,1 kHz | 16-bit | Mono | WAV PCM |
| EASY CHECKOUT | 44,1 kHz | 16-bit | Mono | WAV PCM |
| Cac don vi SCOT di san | 11.025 kHz hoac 22,05 kHz | 16-bit | Mono | WAV PCM |
Thu vien yeu cau tren mot may bay NCR duoc tổ chuc trong mot cau truc thu muc o do moi ten tep WAV tuong ung voi mot ma su kien yeu cau trong cau hinh phan mem POS. Các quy uoc dat ten khac nhau theo tuy chinh nha ban le — một triên khai Kroger co the su dung cac ma yeu cau khac voi mot triên khai Walmart thậm chí tren phan cung NCR giong het.
Tro ngan tao va han che:
Cac he thong loa tu dong thanh toan NCR la cac trai xe tai 3-5 watt trong mot tran kep, phan mem dong goi. Ho khong phai la cac lo a high-fidelity. Cac yeu cau qua to lam cho anh vao; quá coi dep yeu cau khong co dieu chinh. Muc tieu -18 LUFS tich hop voi tran dien tro -3 dBTP (dung dai) cho tieu thuc am luong.
Tu dong Thanh toan Diebold Nixdorf: Cac He thong BEETLE va TP Application
Diebold Nixdorf (truoc day la Wincor Nixdorf) tao ra cac dong tu dong thanh toan BEETLE va TP Application duoc tim thay chinh yeu trong cac chuoi sieu thi Chau Au, bao gom cac hoat dong Chau Au cua Carrefour, va trong mot so nha ban hang chuyen biet My. Kien truc cua ho tuong tu nhu cua NCR nhung voi nhung uu tien dinh dang audio khac nhau.
Cac he thong BEETLE POS chay Windows va su dung nen tang ung dung Storelogix hoac ProFIT cua Diebold Nixdorf. Cac yeu cau audio duoc load duoi dang cac tep WAV vao mot thu vien phuong tien tren may bay. Cac he thong BEETLE tao ra hien tai chap nhan WAV mono 16-bit 44,1 kHz; cac don vi di san thong thuong yeu cau 11.025 kHz hoac 22,05 kHz.
Cac may bay TP Application (TP6 va TP7 dong) su dung cung mot he thong thu vien yeu cau dua tren WAV. Dong san pham TP7, pho bien trong cac chuoi sieu thi Chau Au luong giao thong cao, ho tro am thanh 44,1 kHz tren firmware hien tai.
Tieu thuc audio cho cac he thong Diebold Nixdorf:
| He thong | Ty le Mau | Sau khi Bit | Kenh | Dinh dang |
|---|---|---|---|---|
| BEETLE POS (hien tai) | 44,1 kHz | 16-bit | Mono | WAV PCM |
| BEETLE POS (di san) | 11.025-22,05 kHz | 16-bit | Mono | WAV PCM |
| TP6 Application | 22,05 kHz hoac 44,1 kHz | 16-bit | Mono | WAV PCM |
| TP7 Application | 44,1 kHz | 16-bit | Mono | WAV PCM |
Ghi chu rieng biet Carrefour: Cac triên khai tu dong thanh toan Chau Au cua Carrefour chay Phap va Tieng Anh (cho cac vi tri qua du lich) hoac Phap va Arab (cho cac cua hang Bac Phi). Thu vien yeu cau tren moi may bay chua hai bo ngon ngu voi mot yeu cau chon ngon ngu o dau moi giao dich. Cac he thong TP Application cua Diebold Nixdorf xu ly dieu nay thong qua logic chon ngon ngu trong cau hinh Storelogix, chứ khong phai boi cach doi sao WAV thu muc — toan bo thu vien da ngon ngu ton tai tren moi may bay.
Xay dung Nhan vat Giong noi Tu dong Thanh toan
Mot nhan vat giong noi tu dong thanh toan co nhieu hon mot ban ghi giong noi — no la mot quyet dinh thiet ke am thanh co y tu hinh hinh khach hang nhan thuc mot thuong hieu o thoi diem thanh toan.
Hau het cac nha ban le lon chon giong noi trong dang ky trung lap-trang-am: khong phai lanh hoac giong co-bot (tao mat mo o mot thoi diem suc ep), khong phai qua am ap hoac tao thai (cam thay khong phu hop trong mot boi canh giao dich). Lua chon gioi tinh khac nhau tuy thuoc nha ban le va thi truong — cac chuoi ban le My co lich su thich giong noi nu; mot so chuoi Chau Au su dung giong noi nam; cac triên khai hien dai thuong cung cap ca hai va de cho may bay phat hien tien pho nguoi dung va phuc vu mot giong noi tuong ung.
Cac tinh nang nhan vat giong noi de xac dinh truoc khi tao ra:
- Dang ky giong: Nu, nam hoac trung lap gioi tinh (cai sau ay pho bien ngay cang)
- Da noi: Trung lap Tieng Anh chung cho cac chuoi My; Received Pronunciation hoac trung lap khu vuc cho Vương quốc Anh; cac da tieu chuan quoc gia cho cac thi truong khong phai tieng Anh
- Ty le noi: 130-145 tu tren phut cho cac yeu cau chi dan; nhanh hon chut (150 WPM) cho cac tin nhan xac nhan
- **Da: Am ap nhung khai bao — khong phai cau hoi hoac xin loi (“vui long lam X” hon la “co the ban vui long co the X?”)
- Tinh nhat quan prosodic: Moi clip phai co am luong giong het nhau, cau ca tuong tu va khong co su khac biet co the nghe thay trong am thanh phong tro trong cac clip
Van de tinh nhat quan o quy mo chuoi:
Mot ho so giong noi AI duy nhat giai quyet van de tinh nhat quan theo dinh nghia. Moi yeu cau, bat ke la khi no duoc tao hoac ai da sua ky ban, deu den tu cung mot mo hinh giong noi voi cung cac cai dat. Doi voi mot chuoi phat trien tu 300 den 1.000 cua hang trong ba nam, cac triên khai may bay moi o nam ba nghe giong het voi cac triên khai ban dau o nam 1.
Nay la ly do tai sao kloning giong noi dua tren AI la khong nang gia tri cao nhat cho am thanh ban le doanh nghiep. Xac dinh giong noi mot lan — co the boi kloning tu mot ban ghi dien vien giong noi chat luong cao ma thuong hieu da so huu — sau do tao cac yeu cau khong han tro tu ho so duoc kloning do tai vo hang.
Viet Cac Ky Ban Yeu cau Tu dong Thanh toan cho Dau ra Giong noi AI Tu nhien
Ky ban la noi tao ra cac du an giong noi kios DIY hau het tao ra cac ket qua xau. Yeu cau tu dong thanh toan co mot cau truc ngon ngu dac biet khac voi TTS hoi thoai.
Giu cac yeu cau ngan va buoc ra lenh. “Vui long dat hang hoa vao khu vuc dong goi” (7 tu) dung. “Co the ban vui long dam bao dat hang hoa cua ban tren da co theo khu vuc dong goi?” sai ca chat luong TTS va trai nghiem nguoi dung. Cac nghien cuu UX ban hang nhat quan cho thay rang cac yeu cau ngan hon giam thoi gian giao dich va su nhap nhang cua ng mua sach.
Su dung dau cam nhu dieu khien prosody. Mot dau phay tao mot tam dung ngan trong hau het cac tao giong noi AI. “Chao mung. Vui long quet san pham dau tien cua ban.” tao mot thoi ket cau sach. Khong co chu ky, “Chao mung vui long quet san pham dau tien cua ban” chay cung nhau va nghe khong tu nhien.
Tranh doc so khong ro rang. Viet “bon dong va nam muoi xu” khong phai “$4,50” — mot so he thong TTS doc cai sau nhu “dong 4 diem 50” hoac “bon diem nam khong dong.” Hay ro rang ve cach ban muon cac so duoc doc, dac biet la cho cac tieu chi, so luong va so gang.
Cac ky ban xac minh tuoi yeu cau ro rang tren tat ca. Cac yeu cau nay kich hoat cac quy trinh tuân thủ. Cac ky ban nhu “Mot co van cua hang phai xac minh tuoi cua ban cho san pham nay. Vui long cho.” phai ro rang, co quyen lap lenh va mien trong ng on gam xuat du lieu co the lam cho yeu cau nghe lua.
Cac danh muc thu vien yeu cau tu dong thanh toan tieu chuan:
| Danh muc | Ky ban Vi du | So luong Dien hinh |
|---|---|---|
| Chao va quet | ”Chao mung. Vui long quet san pham dau tien cua ban.” | 3-5 |
| Khu vuc dong goi | ”Vui long dat hang hoa vao khu vuc dong goi.” / “Hang hoa khong mong doi trong khu vuc dong goi.” | 8-12 |
| Canh bao can bang | ”Vui long lay ra tat ca hang hoa tu khu vuc dong goi.” / “Hang hoa da bi loai — vui long quet lai.” | 4-6 |
| Yeu cau thanh toan | ”Vui long chon phuong thuc thanh toan.” / “Vui long chen the cua ban.” / “Vui long chum the cua ban.” | 10-15 |
| Loyalitas va phieu giam gia | ”Co the ban co the loyalitas hoac phieu giam gia khong?” / “The loyalitas duoc chap nhan.” | 4-6 |
| Kiem chung tuoi | ”San pham nay yeu cau kiem chung tuoi. Mot thanh vien doi se ho tro ban.” | 2-3 |
| Loi va ghi de | ”Vui long cho tro giup.” / “Mot thanh vien doi da duoc thong bao.” | 5-8 |
| Giao dich hoan thanh | ”Giao dich duoc phe duyet. Vui long lay hoa don cua ban.” | 3-4 |
| Rieng co hang | Cac loi chao theo mua, cac thong diep khuyen mai, ten cua hang trong yeu cau mo dau | 5-20 |
| Dong cua/dang cho | ”Chao mung toi [ten cua hang]. Vui long quet san pham dau tien cua ban khi san sang.” | 2-4 |
Tong cong tren moi ngon ngu: thong thuong 80-150 clip cho mot thu vien lajur don toan phan.
Tuan thu WCAG 2.1 De truy cap cho Giong noi Kios Retail
Cac may bay tu dong thanh toan la cac co so cong cong theo ADA o My va theo nhung luat tu do tuong duong o EU (Dieu le De truy cap chau Au, co hieu luc thang 6 nam 2025 cho cac giao dien digital ban le) va Vương quốc Anh. WCAG 2.1 cung cap tieu chuan ky thuat ma hau het cac kiem toan de truy cap su dung de danh gia am thanh kios.
Cac Tieu chi Thanh cong WCAG 2.1 Phu hop cho am thanh tu dong thanh toan:
1.1.1 Noi dung Khong phai van ban (Cap A): Moi yeu cau hinh anh tren man hinh kios phai co mot dieu chinh audio. Neu man hinh hien thi “dat hang hoa vao khu vuc dong goi” duoi dang mot lo tro truoc mat, yeu cau audio cung phai phat. Cac yeu cau chi hinh anh khong tuan thu tieu chi nay.
1.3.3 Dac tinh Cam giac (Cap A): Cac chi dan khong duoc chi tro tuy vao cac dac tinh hinh anh. “Nhan nut xanh” ma khong co mot chi dan audio tuong duong khong tuan thu; “Nhan nut xanh co nhan OK” voi mot dieu chinh audio tuan thu.
1.4.2 Dieu khien Audio (Cap A): Neu audio phat tu dong hon 3 giay, nguoi dung phai co the tam dung, dung hoac dieu khien am luong. Tren mot may bay tu dong thanh toan, dieu nay thong thuong duoc dap ung boi viec cung cap mot nut dieu khien am luong tren giao dien man hinh cam ung.
1.4.3 Luong Dieu khien (doi voi van ban tren man hinh, Cap AA): Khong phai audio dac biet, nhung co lien quan den UI kios tich hop dong hành yeu cau giong noi.
2.4.6 Tieu de va Nhan (Cap AA): Gan den man hinh doc — ap dung khi kios pho lo mot lo tro jack tai nghe cho tinh nang nghe tu nhan, tro co WCAG tuân thủ ATM tuong tu.
Cac yeu cau tao ra de truy cap thuc te:
- Roai noi toi thieu: dau ra giong noi AI phai da duoc diem tren 90% tren cac bai kiem tra kha nang hoc tu (Modified Rhyme Test hoac tuong tu) thong qua lo a tren bo duoc cua may bay o nhung am luong khoang khong 65 dB SPL
- Ty le noi: 120-150 WPM cho cac yeu cau chi dan; di nhanh hon lam giam de hieu duoc cho cac khach hang co su khac biet xu ly nhan thuc
- Am luong: -18 LUFS tich hop nhat quan tren tat ca cac clip; su thay doi am luong giua cac yeu cau hoa loan cac nguoi dung co tro nghe
- Cang thi nghe tu nhan: Kios giao thong cao voi mot lo tro jack tai nghe phai tao am thanh sach tren cac do tan suat lo tro tieu chuan — mot muc tieu am luong khac voi dau ra lo a
De xem tham khao va hon cao ve tinh phu hop de truy cap giong noi AI cho cac dao nhat hang cong, huong dan cua chung toi ve tao giong noi AI cho cac yeu cau lobby ATM bao gom cac yeu cau ADA va WCAG lap chong cho cac kios tai chinh, co tran anh day du hoa nan de truy cap.
Giong noi Tu dong Thanh toan Da ngon ngu: Cac Mo hinh Walmart, Kroger, Carrefour
Ba chuoi ban le nhan tho nhat triên khai AI giong noi tu dong thanh toan da ngon ngu dai dien ba cach tien hanh khac nhau de cau hanh thách thuc da ngon ngu.
Walmart My: Tieng Anh + Tay Ban Nha
Cac may bay tu dong thanh toan Walmart My trong cac thi truong co dan so Hispanic cao cung cap cac bo yeu cau Tieng Anh va Tay Ban Nha. Su lua chon ngon ngu xay ra duoi dang au dau tien cua giao dich (mot yeu cau “Chon ngon ngu” voi nut man hinh cam ung) hoac thong qua mot kieu thich pho bien ngon ngu gan voi tai khoan loyalitas cua khach hang.
Tinh nhan vat giong noi Walmart cho Tieng Anh la mot giong noi nu Tieng Anh My trung lap — mot trong cac giong noi tu dong thanh toan duoc nhan ra nhat trong ban le My. Ban tho Tay Ban Nha duy tri mot gia tu tuong tu nhung voi mot da Tay Ban Nha Latin My trung lap (tranh su dac trung hang hoc co the khong bao gom cac nguoi noi tu lat Tay Ban Nha tu cac nhan thuc nen tang khac).
Thuc hien ky thuat: Tren cac may bay FastLane NCR tai Walmart, hai thu vien ngon ngu duoc luu tru trong cac thu muc rieng le (vi du, /prompts/en/ va /prompts/es/) va ung dung POS chuyen doi cac duong dan thu muc dua tren co tieu chuan thich ngon ngu duoc dat o dau phien.
Kroger My: Tieng Anh + Nhung xem xet Khu vuc
Cac triên khai tu dong thanh toan Kroger tren cac co cau (King Soopers, Fred Meyer, Ralphs, Harris Teeter) su dung Tieng Anh la ngon ngu chinh voi nhieu tro giup Tay Ban Nha o cac thi truong phu hop. Phuong phap cua Kroger co lich su nhan man mot da giong noi am ap va hoi thoai hon Walmart — phan anh theo vi tri thi truong ban hang cong dong cua thuong hieu.
“Giong noi Kroger” tren toan bo mang lu,c tu dong thanh toan du la kho nhan biet du de chuoi da dau tu vao su nhat quan giong noi nhu mot khac biet hoa thuong hieu — chinh xac la tro su dung ma ho tro kloning giong noi AI bang cach cho phep mot thuong hieu so huu va lua da mot nhan vat giong noi dac biet.
Carrefour: Phap, Arab va Cac ngon ngu Dac biet Thi truong
Carrefour hoat dong tren 35+ nuoc voi cac triên khai tu dong thanh toan yeu cau cac thu vien yeu cau that su da ngon ngu. Phap la ngon ngu co ban; Arab la ngon ngu thú hai cho cac thi truong Bac Phi (Maroko, Tunisia, Aljazair, Ai Cap); Tay Ban Nha duoc su dung o Tay Ban Nha va cac phan cua My Latinh.
Phuc tap ky thuat o Carrefour co y nghia: mot may bay Diebold Nixdorf TP7 duy nhat trong mot Carrefour Morocco co the yeu cau Phap va Arab Morocco (Darija) hoac Phap va Arab Tieu chuan Hien dai (MSA) tuy thuoc vao tap dung khach hang - va hai bien the Arab du la khac biet nhat de cac thu vien yeu cau rieng le duoc yeu cau.
Tao giong noi AI tro cap dieu nay bang cach cho phep doi audio Carrefour tao cac thu vien bien the Arab kho trong cung mot ky ban yeu cau ma khong can thue tai nang rieng biet cho Darija va MSA.
Kien truc Tro ngan Ngon ngu
Hai phuong phap tru thong cho kien truc audio kios da ngon ngu la:
| Phuong phap | Cach hoat dong | Tot nhat cho |
|---|---|---|
| Lua chon ngon ngu o dau phien | Khach hang chon ngon ngu tren man hinh dau; phien phat tu thu vien ngon ngu do | Cac cua hang da dang; tien pho nguoi dung ro rang |
| Tien pho loyalitas pho bien | Ngon ngu gan vao tai khoan loyalitas; tu dong chon tren quay the | Cac khach hang thuong xuyen; giam ma sat cho khach hang da biet |
| Am thanh song song (ca hai ngon ngu) | Tao mot clip hop nhat tren moi yeu cau: Tieng Anh + tam dung + Tay Ban Nha | Dieu khien di san khong the chuyen doi cac thu muc tro giup trong phien |
| TTS Dong | TTS tren thiet bi hoac tren co so API tao ra moi yeu cau phat truc tiep | Tinh linh hoat cao nhat; yeu cau mot dong co TTS latency thap va truy cap mang |
Doi voi mot boi canh triên khai gan — giong noi AI tao ra o cac lajur dat hang drive-through, o dau cac yeu cau da ngon ngu phuc vu cac khach hang chua tien hanh chon mot ngon ngu — xem huong dan cua chung toi ve tao giong noi AI cho cac dat hang drive-thru, bao gom logic phat hien va thay doi dong cua lo a ngoai troi.
Quy trinh tao Ky thuat: Xay dung Thu vien Yeu cau Ban le
Day la quy trinh tao de tao ra mot thu vien yeu cau tu dong thanh toan toan phan su dung mot tao giong noi AI:
Buoc 1 — Kiem toan tieu thuc phan cung. Yeu cau tai lieu tich hop audio tu ky su thong thong NCR Voyix hoac Diebold Nixdorf. Lay ty le mau duoc yeu cau, sau khi bit, yeu cau mono/stereo, codec (luon la WAV PCM cho cac he thong nay), va quy uoc dat ten tep cho thu muc thu vien yeu cau.
Buoc 2 — Soan thao ky ban yeu cau toan phan. Danh sach moi ma su kien co the phat sinh ung dung POS. Hau het cac triên khai NCR va Diebold Nixdorf dung voi mot thu vien yeu cau co ban tu OEM — lay dieu nay duoi dang tham khao. Them cac yeu cau dac biet nha ban le (ten cua hang, chuong trinh loyalitas, ten phuong thuc thanh toan nhan hang rieng).
Buoc 3 — Xac dinh cac tham so nhan vat giong noi. Dat dang ky giong, ty le noi (130-145 WPM cho cac yeu cau chi dan), da va da. Neu gop y mot giong noi thuong hieu co san, dem theo mot mau ghi am tham khao de kloning giong noi.
Buoc 4 — Tao dot. Nhap danh sach ky ban yeu cau toan phan, chon ho so giong noi, dat dinh dang dau ra tren tieu thuc. Xu ly tat ca cac clip trong mot dot de dam bao cac cai dat giong noi nhat quan tren moi tep. Khong tao cac clip trong cac phien rieng le voi cac cai dat khac — su thay doi am luong va prosody giua cac clip co the nghe thay trong tao ra.
Buoc 5 — Chuan hoa am luong. Muc tieu -18 LUFS tich hop voi tran dien tro -3 dBTP. Ap dung cho moi clip trong dot. Cac cong cu: Loudnorm trong FFmpeg, hoac mot chuan hoa am luong dac biet. Khong su dung chuan hoa diem dau — no tao am luong co nhan thuc khong nhat quan.
Buoc 6 — Them cac dam dung kesunyia. Cong 50-100ms kesunyia; them 200ms kesunyia. Hau het cac dieu khien kios cat dau audio ma khong co dam dung kesunyia ngan. Kesunyia theo sau ngan chan su tao mat va khi yeu cau tiep theo duoc kich hoat.
Buoc 7 — Doi ten cho cac ma yeu cau. Doi ten tep tren quy uoc dat ten dieu khien. Mot ket qua giua ten tep va ma su kien duy buoc co nghia yeu cau phat kesunyia — che do loi pho bien nhat trong cac triên khai thu vien yeu cau tuy chinh.
Buoc 8 — Kiem toan xac minh. Triên khai thu vien yeu cau den mot may bay kiem toan. Di qua mot truot thao tac giao dich toan phan bao gom cac trang thai loi (khong khop khu vuc dong goi, tu choi the, kich hoat xac minh tuoi). Xac minh moi yeu cau duoc phat dung, o thoi diem dung, o am luong dung.
Buoc 9 — Tai lieu cac cai dat ho so giong noi. Luu moi tham so duoc su dung: mo hinh giong noi, ty le noi, cai dat am luong, dinh dang dau ra. Khi cap nhat ky ban yeu cau tao lai mot clip sau sau nam, da neu cai dat ban dau dam bao clip moi nghe giong het voi thu vien co san.
De xem boi canh ve cach lien quan logic tao dot tuong tu ap dung cho cac yeu cau suara may ban hang — mot tro su dung lo a kios tuong tu nhung don gian hon — xem huong dan cua chung toi ve tao giong noi AI cho cac may phan phoi tu dong.
So sanh Cac nen tang Giong noi AI cho Tao Kios Ban le
| Nen tang | Xuat WAV | Ky ban Dot | Kloning Giong noi | Offline | Tro giup SSML |
|---|---|---|---|---|---|
| ElevenLabs | Co (tra phi) | Thong qua API | Co (tra phi) | Khong | Han che |
| Murf | Co (tra phi) | Thong qua API | Han che | Khong | Co |
| Azure TTS | Co | Co (SSML) | Giong noi Than Kinh Tuy chinh | Khong | Toan bo |
| Google Cloud TTS | Co | Co | Giong noi Tuy chinh | Khong | Toan bo |
| VoxBooster | Co | Co | Co (dia phuong) | Co (Windows) | Co |
Tieu chi chinh cho triên khai ban le:
Xu ly ngoai tuyen/dia phuong: Cac may bay kios trong nhung moi truong back-of-house ban le co the co tro cap internet khoang han che vi cac ly do tuân thủ PCI-DSS. Mot tao giong noi dia phuong chay tren may tao san xuat ma khong co cac goi API cloud loai bo mot cuoc tro chuyen tuân thủ.
Kloning giong noi tu ban ghi tham khao: Neu mot nha ban le da co mot ban ghi bao ta giong noi hien co xac dinh giong noi thuong hieu cua ho, kloning tham khao do — hon la chon mot giong noi chung co san — bao ton tong suc thuong hieu. Giong noi duoc kloning tao tat ca cac yeu cau moi va cap nhat tai vo hang tu cung mot nhan dung giong noi.
Xuat dot voi cac cai dat nhat quan: Tao 120 clip mot buoc mot thong qua mot giao dien web khong thuc te. Xu ly dot tu mot tep ky ban voi cac cai dat giong noi la cam bao dam bao moi clip trong thu vien nhat quan.
SSML de dieu khien cach doc: Cac yeu cau ban le thuong bao gom cac ma san pham, dinh dang gia, va cac ten chuong trinh loyalitas ma cac dong co TTS co the doc khong mong doi. SSML cho phep ban chi dinh cach doc mot cach ro rang: <say-as interpret-as="currency">$4,50</say-as> hoac <say-as interpret-as="cardinal">4</say-as> san pham.
Doi voi cac quy trinh kloning giong noi — dac biet la da neu ban ghi giong noi thuong hieu hien co — huong dan kloning giong noi cua chung toi cho cac con duong phat am bao gom phuong phap, tieu chuan chat luong va cac yeu cau ky thuat cho kloning muc do tao ra.
Nhung Sai lam Pho bien trong Tao Giong noi Kios Ban le
Tao theo che do stereo. Moi dieu khien tu dong thanh toan lon — NCR, Diebold Nixdorf va hau het cac OEM phu — yeu cau WAV mono. Cac tep stereo hoac bi tu choi hoac duoc phat sai. Tao mono tu dau; khong tin tuo vao dieu khien de downmix.
Su dung truc tiep cac giong noi TTS nguoi tieu dung ma khong co chuan hoa am luong. Cac nen tang TTS nguoi tieu dung toi uu hoa cho phat hang qua tai nghe hoac lo a o khoang -14 LUFS. Cac lo a kios ban le la cac moi truong am thanh khac nhau. Ma khong co chuan hoa am luong thanh -18 LUFS, cac yeu cau se rat to khong nhat quan tren toan bo thu vien.
Bo qua dam dung kesunyia da dau. Cac dieu khien kich hoat am thanh ngay lap tuc tren dieu kien loi se cat suku dau tien cua mot yeu cau bat dau tai mau khong. Mot dam dung kesunyia 50-100ms ngan chan dieu nay.
Cac cai dat giong noi khac nhau giua cac phien cap nhat. Tao thu vien ban dau vao thang 1 va cap nhat ba yeu cau vao thang 9 voi cac cai dat pitch hoac toc do hoi khac tao ra khong nhat quan co the nghe thay trong tao ra. Khoa va tai lieu cac cai dat ngay thu nhat.
Ngon ngu mem trong cac yeu cau tuân thủ. Xac minh tuoi va cac yeu cau kiem chung ID ton tai de tuân thủ phap ly. Lam yeu cac yeu cau do (“co the ban can phai cho chi tiet”) tao ra su mo ho tro khoang do chuyen bay bao tai qua ban hang va co the tao ra chiếc trách nhiệm. Cac yeu cau nay can phai ro rang, tru tiep va rõ ràng.
Bo qua vong idle/chao mung. Yeu cau keadaan idle duoc phat khi may bay doi cho mot khach hang la mot trong nhung am thanh duoc nghe nhat tren cua hang. Cac da cua no xac lap co man dau tien cua trai nghiem thanh toan. Khong xem dieu nay nhu mot dieu phu.
Doi voi cac tao giong noi huong den cac tao noi dung hon la cac triên khai ban le doanh nghiep, huong dan voice changer cua chung toi cho cac tao noi dung bao gom cac yeu cau chat luong va quy trinh lam viec khac nhau cho cac tro su dung streaming va phuong tien xã hội.
Cau hoi Thuong gap
AI giong noi tu dong thanh toan la gi?
AI giong noi tu dong thanh toan la mot he thong text-to-speech nhung trong kios ban le huong dan khach hang thong qua quy trinh scan-and-pay. No tao ra cac yeu cau ban nhe tai cac lajur tu dong thanh toan Walmart, Kroger va Carrefour — “Vui long dat hang hoa vao khu vuc dong goi”, “Hang hoa khong mong doi trong khu vuc dong goi”, “Vui long chen the cua ban” — su dung mot nhan vat giong noi tong hop phu hop o moi may trong mot chuoi cua hang.
Phan cung nao chay cac yeu cau giong noi tu dong thanh toan tai cac nha ban le lon?
NCR Voyix (truoc day la NCR) va Diebold Nixdorf la hai nha san xuat OEM tu dong thanh toan tru thong. Cac dong FastLane va SelfServ cua NCR phat hang qua mot loa onboard do mot dieu khien dua tren Windows hoac Linux. Cac he thong BEETLE va TP Application cua Diebold Nixdorf su dung mot kien truc tuong tu. Ca hai deu chap nhan cac tep audio WAV duoc load vao thu vien yeu cau tren dieu khien — tao giong noi AI tao ra cac tep WAV do.
Toi co the tao mot giong noi tu dong thanh toan phu hop voi WCAG 2.1 nhu the nao?
Tieu chi Thanh cong WCAG 2.1 1.4.2 (Dieu khien Audio) va 1.3.3 (Dac tinh Cam giac) la cac diem tham chieu phu hop nhat. Trong thuc te: moi yeu cau hinh anh phai co mot yeu cau audio tuong duong, audio khong duoc tu dong phat qua 3 giay ma khong co dieu khien nguoi dung, va giong noi phai de hieu duoc o nhung am luong kios binh thuong — thong thuong 65-75 dB SPL o 0,5 m. Su dung mot da noi trung lap o 130-150 WPM va do loud phu hop (-18 LUFS hop nhat).
Co the mot giong noi AI bao phu mot kios tu dong thanh toan da ngon ngu khong?
Mot dong co AI duy nhat co the tao ra cac yeu cau nhieu ngon ngu tu cung mot pro-fil giong noi, nhung nhan vat giong noi dau ra se khac nhau theo ngon ngu vi moi mo hinh ngon ngu duoc dao tao tren cac mau pha-bieu ban dia phuong. De duy tri tinh nhat quan cua thuong hieu tren cac ngon ngu, xac dinh mot san ki chinh thuc (nuong chieu, trung lap, hoi chinh thuc) va danh gia dau ra cua moi ngon ngu trong do truoc khi trien khai. Cac cua hang Walmart US thong thuong chay Tieng Anh + Tay Ban Nha; Carrefour Phap chay Phap + Arab cho cac vi tri luu luong cao.
Dinh dang audio nao ma kios NCR Voyix va Diebold Nixdorf chap nhan?
Hau het cac he thong tu dong thanh toan NCR Voyix chap nhan WAV PCM 16-bit o 22,05 kHz hoac 44,1 kHz mono. Cac dong BEETLE va TP Application cua Diebold Nixdorf thong thuong su dung WAV mono 16-bit o 11.025 kHz hoac 22,05 kHz cho thu vien yeu cau di san va 44,1 kHz cho cac he thong tao ra hien tai. Luon yeu cau tieu thuc tich hop audio tu ky su thong thong — sai so dinh dang la ly do pho bien nhat cac yeu cau giong noi tuy chinh khong phat.
Mot kios tu dong thanh toan dien hinh can bao nhieu yeu cau audio?
Mot thu vien yeu cau tu dong thanh toan tieu chuan cho mot may lajur don nhan chua 80-150 clip WAV rieng le bao gom cac yeu cau quet, canh bao khu vuc dong goi, luong thanh toan, cac yeu cau chuong trinh loyalitas, kiem chung tuoi, phuc hoi loi va cac thong diep rieng cua cua hang. Nhan len toan bo chuoi ban le 500 cua hang voi 4 lajur moi cua hang va 2 ngon ngu, do co the la 1,2 trieu tep audio rieng le — tao ra dot xay dung AI la cach duy nhat thuc te de tao ra va bao quan o quy mo.
Co phai VoxBooster hoat dong cho san xuat giong noi kios ban le khong?
VoxBooster chay tren Windows va tao dau ra WAV chat luong cao voi kloning giong noi AI tuy chinh — co ich cho viec tao nhan vat giong noi thuong hieu phu hop toan bo thu vien yeu cau kios. Quy trinh lam viec phu hop voi nhung gi cac doi audio ban le lam: ghi am hoac kloning mot giong noi tham khao, tao ra tat ca cac yeu cau tu mot danh sach ky ban theo lan, xuat nhu WAV mono o ty le mau dieu chinh. Ban tho gra than bao gom dau ra du de xac thuc chat luong giong noi truoc khi cam ket de chay san xuat thu vien yeu cau day du.
Ket luan
AI giong noi tu dong thanh toan la mot kỹ nang tao ra, khong phai chi la mot lua chon cong nghe. Giong noi “vui long dat hang hoa vao khu vuc dong goi” ma khach hang nghe tai Walmart, Kroger va Carrefour duoc thiet ke va tao ra voi cac yeu cau phan cung dac biet, tieu chuan de truy cap va cac huong dan suara thuong hieu co y tu — va duy tri no tren hang ngan lajur va nhieu ngon ngu yeu cau mot quy trinh lam viec ma ghi am phong thu khong the duy tri o quy mo.
Cac tao giong noi AI dia chi moi chuan: yeu cau phan cung NCR Voyix va Diebold Nixdorf (WAV mono 16-bit o ty le mau dung), tuân thủ WCAG 2.1 de truy cap (am luong nhat quan, ty le noi co the hieu duoc, cac dieu chinh audio cho tat ca cac yeu cau hình anh) va cac phuc tro da ngon ngu (mot cong viec dot tren moi ngon ngu tu cung mot pro-fil giong noi). Quy trinh lam viec — ky ban, tao ra, chuan hoa, dat ten, xac minh — co the lap lai va co the xem xet theo cach ma cac phien hop phong thu ad-hoc khong phai.
VoxBooster xu ly tao giong noi AI va kloning giong noi tuy chinh tren Windows, lam cho it thuc te de xay dung mot thu vien yeu cau ban le toan phan tu mot nhan vat giong noi thuong hieu duoc xac dinh. Cung mot quy trinh dia phuong va ngoai tuyen teo tao it va truong hop dieu chinh chat luong tuân thủ API PCI-DSS kieu nhu cap nhat yeu cau trong chieu chieu hon la dat phong thu trong ba tuan. Tai khoan tho gra than — khong the the tin dung yeu cau.
Tai xuong VoxBooster — tai khoan tho gra than, khong the the tin dung yeu cau.