Dokimantasyon API

Intégrer tradiksyon pwisan nan aplikasyon ou ak nou senp REST API.

Kòmanse

TranslateAPI a bay yon entèfas REST senp pou tradiksyon tèks ant plis pase 180 lang.Tout pwen fen API a bay repons JSON.

URL baz: https://api.translateapi.ai/api/v1/
Enstalasyon rapid

Fè premye demann tradiksyon ou:

curl -X POST https://api.translateapi.ai/api/v1/translate/ \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "text": "Hello, world!",
    "target_language": "es"
  }'
import requests

response = requests.post(
    "https://api.translateapi.ai/api/v1/translate/",
    headers={
        "Authorization": "Bearer YOUR_API_KEY",
        "Content-Type": "application/json"
    },
    json={
        "text": "Hello, world!",
        "target_language": "es"
    }
)

result = response.json()
print(result["translated_text"])  # "Hola, mundo!"
const response = await fetch("https://api.translateapi.ai/api/v1/translate/", {
    method: "POST",
    headers: {
        "Authorization": "Bearer YOUR_API_KEY",
        "Content-Type": "application/json"
    },
    body: JSON.stringify({
        text: "Hello, world!",
        target_language: "es"
    })
});

const result = await response.json();
console.log(result.translated_text);  // "Hola, mundo!"
using System;
using System.Net.Http;
using System.Text;
using System.Threading.Tasks;
using Newtonsoft.Json;

class Program
{
    static async Task Main()
    {
        var client = new HttpClient();
        client.DefaultRequestHeaders.Add(
            "Authorization", "Bearer YOUR_API_KEY"
        );

        var content = new StringContent(
            JsonConvert.SerializeObject(new {
                text = "Hello, world!",
                target_language = "es"
            }),
            Encoding.UTF8,
            "application/json"
        );

        var response = await client.PostAsync(
            "https://api.translateapi.ai/api/v1/translate/",
            content
        );

        var result = await response.Content.ReadAsStringAsync();
        var data = JsonConvert.DeserializeObject<dynamic>(result);
        Console.WriteLine(data.translated_text);  // "Hola, mundo!"
    }
}

Authentification

Authentifier demann ou an lè l sèvi avèk yon kle API. Ou ka kreye kle API soti nan tablèt.

Authentication Header (Recommended)
Authorization: Bearer ta_your_api_key_here
Paramèt kesyon
https://api.translateapi.ai/api/v1/translate/?api_key=ta_your_api_key_here
Kenbe kle API ou an sekirite! Pa ekspoze yo nan kòd sou bò kliyan an oswa repozitwa piblik.

Tradiksyon

Tradiksyon tèks nan yon lang objektif sèl.

POST https://api.translateapi.ai/api/v1/translate/
Kò demann lan
Paramèt Tipe Required Deskripsyon
text string Wi Text to translate (max 50, 000 characters)
target_language string Wi* Target language code (e.g., "es", "fr", "de")
source_language string Non Source language code. Default: "auto" (auto-detect)

* Itilize target_language (string) pou lang sèl oswa target_languages (array) pou plizyè. Gade Tradiksyon Multi- Target.

Reponn
{
    "translated_text": "Hola, mundo!",
    "source_language": "en",
    "target_language": "es",
    "translations": {
        "es": "Hola, mundo!"
    },
    "character_count": 13,
    "translation_time": 0.45
}

Tradiksyon Multi- Target

Tradwi tèks nan plizyè lang nan yon sèl demann. Itilize menm pwen an bout tankou tradiksyon sèl.

Tipe: Ou ka tradui nan jiska 50 lang nan yon sèl demann.
POST https://api.translateapi.ai/api/v1/translate/
Kò demann lan
{
    "text": "Hello, world!",
    "target_languages": ["es", "fr", "de", "ja"],
    "source_language": "en"
}

Itilize target_languages (array) an plas de target_language (string) pou plis pase yon objektif.

Reponn
{
    "source_language": "en",
    "translations": {
        "es": "Hola, mundo!",
        "fr": "Bonjour, monde!",
        "de": "Hallo, Welt!",
        "ja": "こんにちは、世界！"
    },
    "character_count": 52,
    "translation_time": 2.31
}

Tradiksyon

Tradwi plizyè tèks nan yon fwa ak pwosesis async. Soumèt yon batch ak sondaj pou rezilta yo.

Limit: maksimòm 500 tèks pa batch, maksimòm 750 atik total (tèks × lang target).Jobs timeout 30 minit apre pwosesis la kòmanse (tan tann nan file a pa konte).
Tan pwosesis la varye selon lang la: lang komen (espanyòl, franse, alman, elatriye) sèvi ak modèl vit (~0.1s pou chak tèks), pandan ke lang ki mwens komen yo sèvi ak modèl multilengwistik nou an (~1-3s pou chak tèks). Yon batch 100 tèks anjeneral fini nan 10-30 segonn pou lang komen, oswa 2-5 minit pou sa ki mwens komen. Pou pi bon rezilta, voye 1 lang target pou chak demann batch ak kenbe gwosè batch la anba 50 tèks.
POST https://api.translateapi.ai/api/v1/translate/batch/
Etape 1: Soumèt batch
curl -X POST https://api.translateapi.ai/api/v1/translate/batch/ \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "texts": ["Hello", "Goodbye", "Thank you"],
    "target_language": "es",
    "source_language": "en"
}'
Reponn (HTTP 202 aksepte)
{
    "job_id": "67535b2b-c9e3-4f82-9499-e237edbc1dd8",
    "status": "pending",
    "total_texts": 3,
    "queue_position": 1,
    "source_language": "en",
    "target_languages": ["es"],
    "character_count": 22,
    "credits_remaining": -1,
    "poll_url": "https://api.translateapi.ai/api/v1/jobs/67535b2b-c9e3-4f82-9499-e237edbc1dd8/"
}
Etape 2: Sondaj pou rezilta yo
GET https://api.translateapi.ai/api/v1/jobs/{job_id}/
Polling Example (Python)
import time, requests

job_id = response.json()["job_id"]
total = response.json()["total_texts"]
headers = {"Authorization": "Bearer YOUR_API_KEY"}

print(f"Batch submitted: {total} texts (job {job_id})")

while True:
    result = requests.get(f"https://api.translateapi.ai/api/v1/jobs/{job_id}/", headers=headers).json()
    status = result["status"]
    processed = result.get("processed_texts", 0)
    progress = result.get("progress_percentage", 0)

    if status == "completed":
        print(f"Completed: {processed}/{total} texts in {result.get('processing_time', 0):.1f}s")
        translations = result["result_data"]["translations"]
        break
    elif status == "failed":
        print(f"Failed at {processed}/{total}: {result.get('error_message', 'unknown')}")
        raise Exception(result.get("error_message", "Translation failed"))
    elif status == "pending":
        queue_pos = result.get("queue_position", "?")
        print(f"Queued (position {queue_pos}) — waiting for GPU worker...")
    else:
        print(f"[{status}] {processed}/{total} ({progress:.0f}%)")

    time.sleep(3)
Reponn (anndan — nan liy, ap tann pou GPU)
{
    "job_id": "67535b2b-...",
    "status": "pending",
    "processed_texts": 0,
    "total_texts": 3,
    "progress_percentage": 0.0,
    "queue_position": 3
}
Reponn ( pandan pwosesis la)
{
    "job_id": "67535b2b-...",
    "status": "processing",
    "processed_texts": 1,
    "total_texts": 3,
    "progress_percentage": 33.33,
    "queue_position": null
}
Reponn (komple)
{
    "job_id": "67535b2b-...",
    "status": "completed",
    "processed_texts": 3,
    "total_texts": 3,
    "progress_percentage": 100.0,
    "processing_time": 10.65,
    "result_data": {
        "translations": ["Hola", "Adiós", "Gracias"],
        "source_language": "en",
        "target_language": "es",
        "character_count": 22,
        "processing_time": 10.65
    }
}
Tracking Progress-Real-Time

Tout repons sondaj gen ladan jaden pwogrè tan reyèl konsa ou ka swiv egzakteman sa ki rive ak batch ou:

Fòma Deskripsyon
status Eta travay la: pending (nan liy, ap tann yon travayè GPU), processing (tradiksyon aktif), completed, failed
processed_texts Nimewo tradiksyon endividyèl ki te fèt jouk kounye a. Mizajou nan tan reyèl jan chak tèks tradwi.
total_texts Nombre total de traductions dans ce lot (textes × langues cibles).
progress_percentage Pousantaj fini (0-100). Kalkile soti nan processed_texts / total_texts.
queue_position Pozisyon ou nan liy lan lè estati a se "anndan" (1 = pwochen). Null lè pwosesis la fini. Itilize sa a pou estime tan tann lan epi montre estati liy lan bay itilizatè ou yo.
processing_time Tan total pwosesis la an segonn (disponib lè li fini).
Tipe: Konbyen tan status se "pending", travayè GPU yo ap travay sou lòt lots. Check queue_position Klike sou 1 pou wè konbyen travay ki devan ou (1 = ou se pwochen). Ou pral kòmanse travay ou otomatikman — pa gen okenn aksyon ki nesesè, jis kontinye sondaj la.
Pi bon pratik pou gwo chaj travay
  • Send 1 target language per batch request. This keeps each batch fast and makes progress easy to track.
  • Keep lots at 50-100 texts. Smaller lots complete faster and give you more frequent progress updates.
  • GPU a processes 2 lots in parallel — additional jobs queue and will not start faster.
  • Lè tan pase, re-poll menm job_id la an plas de voye yon nouvo batch. Travay orijinal la ka toujou ap pwosesman sou GPU a.
  • Sondaj chak 3-5 segonn. Sondaj ki pi souvan pa ralanti pwosesis la.
Multi- lang batch

Tradwi plizyè tèks nan plizyè lang nan yon fwa:

{
    "texts": ["Hello", "Goodbye"],
    "target_languages": ["es", "fr"],
    "source_language": "en"
}
Completed result_data
{
    "translations": [
        {"es": "Hola", "fr": "Bonjour"},
        {"es": "Adiós", "fr": "Au revoir"}
    ],
    "source_language": "en",
    "target_languages": ["es", "fr"],
    "character_count": 24,
    "processing_time": 2.45
}
Paramèt demann
Paramèt Tipe Required Deskripsyon
texts array Wi Array of strings to translate
target_language string Wi* Kode lang target pou lang sèl
target_languages array Wi* Array of target language codes for multiple languages
source_language string Non Source language code. Default: "auto"

* Fòmilè target_language or target_languages, pa tou de.

Pwosesis Async: Batch requests return immediately job_id. Sondaj GET /api/v1/jobs/{job_id}/ jiskaske status se "completed", Lè sa a li result_data Pou tradiksyon. progress_percentage pou swiv pwogrè.

Tradiksyon Dokiman

Tradwi tout dokiman pandan y ap kenbe fòmatasyon. Sopòte plizyè fòma dosye.

POST https://api.translateapi.ai/api/v1/translate/document/
Demand (multipart/form-data)
Paramèt Tipe Required Deskripsyon
file file Wi Dokiman pou tradwi (maksimòm 10MB)
target_language string Wi Target language code (e.g., "es", "fr", "de")
source_language string Non Source language code. Default: "auto" (auto-detect)
Tipi Fichiè Soute
  • .txt - Fichiè tèks
  • .docx - Dokiman Word
  • .pdf - Dokiman PDF (ki gen ladan scanned)
  • .json - Fichye JSON (tradiksyon valè string)
  • .xml - Fichye XML
  • .srt - Fichiè sous-titres
  • .po / .pot - Fichye tradiksyon Gettext
  • .jpg / .jpeg - Imèl JPEG (OCR)
  • .png - PNG imaj (OCR)
  • .tiff / .tif - imaj TIFF (OCR)
  • .bmp - BMP imaj (OCR)
  • .webp - WebP imaj (OCR)
OCR sipò: Fichye imaj ak PDF skannen yo ap trete avèk rekonèt karaktè optik (OCR) pou ekstrè tèks anvan tradiksyon. Pou pi bon rezilta, sèvi ak imaj klè ak rezoudyon wo.
Egzanp (cURL)
# Translate a Word document
curl -X POST https://api.translateapi.ai/api/v1/translate/document/ \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@document.docx" \
  -F "target_language=es" \
  -F "source_language=en"

# Translate text from an image (OCR)
curl -X POST https://api.translateapi.ai/api/v1/translate/document/ \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@scanned_page.jpg" \
  -F "target_language=es" \
  -F "source_language=en"
Fichiè imaj yo ap trete ak OCR pou ekstraksyon tèks anvan tradiksyon. Rezilta tradiksyon an se tounen kòm yon .txt ale.
Reponn
{
    "id": 123,
    "original_filename": "document.docx",
    "file_type": "docx",
    "source_language": "en",
    "target_language": "es",
    "status": "completed",
    "character_count": 5420,
    "translated_file_url": "/media/translated/document_es.docx",
    "created_at": "2024-01-15T10:30:00Z",
    "completed_at": "2024-01-15T10:30:05Z"
}
Valè eta
pending Fichiè a telechaje, ap tann pou l dwe trete
processing Tradiksyon an ap fèt
completed Tradiksyon fini, telechaje disponib
failed Tradiksyon an te pèdi (check error_message)
GET https://api.translateapi.ai/api/v1/translate/document/{id}/

Tcheke estati yon tradiksyon dokiman oswa jwenn URL pou telechaje.

Reponn
{
    "id": 123,
    "original_filename": "document.docx",
    "status": "completed",
    "translated_file_url": "/media/translated/document_es.docx",
    "character_count": 5420
}

Deteksyon lang

Deteksyon lang se bati nan chak demann tradiksyon. Set source_language to "auto" (oswa bliye li) epi lang detekte a pral tounen nan repons lan.

POST https://api.translateapi.ai/api/v1/translate/
Kò demann lan
{
    "text": "Bonjour, comment allez-vous?",
    "target_language": "en"
}
Reponn
{
    "translated_text": "Hello, how are you?",
    "source_language": "fr",
    "target_language": "en",
    "translations": {
        "en": "Hello, how are you?"
    },
    "character_count": 28,
    "translation_time": 0.52
}

The source_language Field nan repons lan montre lang detekte lè deteksyon otomatik se itilize.

Lang ki sipòte

Obtenn lis tout lang ki sipòte yo.

GET https://api.translateapi.ai/api/v1/translate/languages/
Reponn
{
    "count": 186,
    "results": [
        {"iso": "en", "name": "English", "en_label": "English"},
        {"iso": "es", "name": "Español", "en_label": "Spanish"},
        {"iso": "fr", "name": "Français", "en_label": "French"},
        ...
    ]
}

Modèl Tradiksyon

Nou itilize modèl tradiksyon open source ki kouri sou infrastrikti GPU nou an. Tout modèl yo gen lisans komèsyal (Apache 2.0).

Modèl Lang Pi bon pou
Helsinki-NLP/opus-mt 50+ pè lang Lang komen (EN, ES, FR, DE, IT, PT, RU, ZH, JA, etc.)
Google MADLAD-400 400+ lang Lang ra, kouvrefe konplè

API a otomatikman chwazi pi bon modèl la pou pè lang ou. Ou ka opcionellement endike yon engine Paramèt:

Motè Deskripsyon
"auto" Default. Tryes HuggingFace first, falls back to MADLAD-400
"huggingface" Force HuggingFace/MarianMT (pi vit, 50+ lang)
"madlad" Force MADLAD-400 (400+ lang)

Manipilasyon erè

API a itilize kòd estati HTTP estanda pou endike siksè oswa erè.

Kòd Deskripsyon
200 Succès
400 Bad Request - Paramèt ki pa valab
401 Pa otorize - Chèn API ki manke oswa ki pa valab
402 Payement requis - Quota caractères quotidienne dépassée
429 Twòp demann - limit vitès depase
503 Service Unavailable - Translation engine temporarily down
Format repons erè
{
    "error": "daily_limit_exceeded",
    "credits_remaining": 0,
    "daily_limit": 100000
}

Limit

Limit yo varye selon plan an. pri pou plis detay:

Plan Karaktè/Moun Pri
Gratis 250,000 $0 Enskri pou gratis
Starter 2,500,000 $9/mo Abònman
Pro 10,000,000 $29/mo Abònman
Biznis 40,000,000 $79/mo Abònman
Eskalye 125,000,000 $199/mo Abònman

Lè ou depase limit ou, ou pral resevwa yon 429 Too Many Requests pa gen okenn repons jiskaske mwa kap vini an oswa ou mete ajou.

Auto-Scaling Cloud Enfrastrikti

TranslateAPI kouri sou ijans GPU NVIDIA A100 dedikatè ak skalaj orizontal otomatik. Lè demann ogmante, ijans GPU adisyonèl yo lanse nan kèk minit pou kenbe tan repons rapid.Sa vle di ke API nou an ka jere demann konparab pratikman san limit san degradasyon — soti nan yon sèl demann nan milye pa minit.

Ranje paj sa a
Mèsi pou rating ou!
/5 ki baze sou Ratings