AI Briefing
KOSign in

Bilibili Releases Index-Translate, a Multilingual Translation Model Family Based on Qwen3.5

·2026.10.04 16:56

Key point

The family includes specialized models for speech, syllable control, and long-document translation, covering 150 languages.

Details

Bilibili has released Index-Translate, a family of multilingual translation models built on Qwen3.5. The core text models support 150 languages and are designed to follow specific translation instructions, including terminology adherence, formatting constraints, and content preservation.

The family extends beyond basic text translation with three specialized variants:

  • Index-Echo: Generates translated subtitles or speech conditioned on the source speaker's voice.
  • Index-Homura: Adjusts translations to match a specified target syllable count.
  • Index-NativeLong: Handles full-document translation with context maintained across passages.
open-sourcetranslationmultilingualspeech-to-speechqwen3.5bilibiliindex-translate
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

PyTorchKRBilibili Releases Index-Translate, a Translation Model Supporting 150 LanguagesTLDR AIAmazon Releases Strands Decider 2B, an Open-Source Decision Model for Agentic WorkflowsHuggingFace BlogHugging Face Launches Open TTS Leaderboard for Scalable Multilingual and Voice Cloning EvaluationGeekNewsPostHog Releases Jeeves: Reasoning-Enhanced Decision Model Based on Qwen3.5-9BHuggingFace BlogAi2 Open-Sources AstaBrief 8B for Fast, Cost-Effective Research Report GenerationHacker NewsDoGBench: First User-Facing Documentation Benchmark Shows No Local Model Scores Above 50%