Translation engines
Translate every passage, without losing the thread.
Use an on-device engine, a dedicated translation API, or a language model that carries context from earlier passages.
Engine families
Three ways to translate
Pick by privacy, platform, and how much document context the translation needs.
The private reader on your device
Native local translation
Apple Translation serves supported iPhone, iPad, and Mac systems; Google on-device translation serves supported Android devices. Language resources may need to be downloaded first.
Apple or Android devicesServices built specifically for translation
Dedicated translation APIs
Use a translation-focused provider for passage-by-passage document translation. Attha exposes a curated eight-language product surface for these cloud adapters.
Cloud · supported app platformsA translator that remembers the preceding passages
Language-model translation
These adapters translate one block at a time while carrying bounded document context and matching terminology forward.
Cloud · supported app platformsCapability map
What does your document need?
The table describes capabilities exposed by Attha today. It is not an accuracy or quality ranking.
Catalog checked July 2026
Engine
Platforms
Context
Terminology
On device
Credential
On deviceLocal · no API key · platform-specific
Apple TranslationiOS 18+ and macOS 15+; local language resources required2
0
0
5
No API key
Google On-device TranslationAndroid API 23+; local language models required1
0
0
5
No API key
Professional MTAutomatic detection · dedicated APIs · BYO credentials
Alibaba Cloud Machine TranslationDedicated translation API with limited preceding context5
2
0
0
Bring credentials
Context translationRecent pairs · rolling summary · glossary
Zhipu GLM Context TranslationRecent pairs, rolling summary, style, and glossary5
5
5
0
Bring credentials
Platform reach is the number of SDK platforms declared by the provider (1–5). Context: 0 = block only, 2 = limited preceding text, 4 = structured references, 5 = rolling document context. Terminology is 5 only when the adapter accepts glossary entries. On-device is 5 only when text is processed by a native local engine. These values describe the current Attha integration, not upstream quality.
Privacy · credentials
Your credentials stay under your control.
Local engines use downloaded language resources on the device. Cloud requests go from the app to the provider using credentials stored on your device.
On-device enginesApple Translation or Google on-device translation
No API keyProfessional MTSix dedicated cloud translation adapters
Bring credentialsContext translationSix language-model adapters
Bring credentialsProvider availability and supported language pairs can vary by platform, account, and region.
Common questions
Before you choose
Which files can Attha import?
TXT, DOCX, SRT, and VTT. You can also paste text directly.
Does every engine work offline?
No. Apple Translation is local on supported Apple systems, and Google on-device translation is local on supported Android systems. Cloud engines require network access and your own credentials.
Does every engine use document context and terminology?
No. Context-aware language-model engines use recent bilingual passages, a rolling summary, and matching glossary entries. Some dedicated translation engines expose a smaller context surface.