Skip to main content
Gemini is Google’s model family and the most natively multimodal option in the translation catalog: text, images, audio, video, and PDF are all first-class inputs. For Document Orders containing diagrams, screenshots, or scanned pages, that is the difference between localizing the whole document and localizing only the parts that were already text.
Available in Ollang Workflows — all Gemini tiers serve Document Translation, Subtitle Translation, and AI Dubbing. See the Translation catalog.

Models

Choosing a tier

Gemini 4 is Google’s current generation, in three stable tiers. All three carry a 1,048,576-token context window shared between input and output, 64K output by default, and accept text, images, audio, video, and PDF:
  • Gemini 4 Carbon is Google’s strongest reasoning model. Use it for the hardest documents and quality-critical content.
  • Gemini 4 Argon is efficient frontier reasoning — faster than Carbon and, in Google’s words, the ideal choice for most knowledge work. It is the default recommendation within Gemini for most localization.
  • Gemini 4 Hydrogen is the fast, efficient everyday tier for high-volume work.
Gemini 3.7 Flash is the previous-generation stable Flash model, with a 1M-token input context, 64K output, and a March 2026 knowledge cutoff. Gemini 3.1 Pro Preview is the reasoning flagship, with a 1M-token input window and a three-tier thinking control (low, medium, high) that trades latency against reasoning depth. Select it for very large documents, whole content sets, and mixed-media source material that needs reasoning rather than throughput. Gemini 2.5 Pro remains stable and available, with broad multilingual coverage drawn from Google’s data. Gemini 3 Pro Preview has been superseded in Google’s model list. It is retained here for Workflows already benchmarked against it; new Workflows should start from Gemini 4.

Where it fits in a workflow

Two situations make Gemini the right call:
  1. Mixed-media documents. Diagrams, embedded screenshots, scanned pages, and image-heavy PDFs. Text-only models silently skip what they cannot read.
  2. Very long single-context Orders. Whole books, complete documentation sets, or full season subtitle runs where consistency across the entire body is the goal.

Trade-offs

  • Gemini 3.1 Pro is a preview release, so its behavior can change between versions. The Gemini 4 tiers are stable.
  • Preview model IDs change as releases ship — Gemini 3 Pro Preview is already superseded. Re-check model status before standardizing a long-lived Workflow on a preview tier.

Reference