# VITA-MLLM/VITA

✨✨[NeurIPS 2025] VITA-1.5: Towards GPT-4o Level Real-Time Vision and Speech Interaction

Repository: https://github.com/VITA-MLLM/VITA
Canonical: https://ross.abutalabs.com/products/vita
Language: Python
License: NOASSERTION
License Family: other
Topics: large-multimodal-models, multimodal-large-language-models, omni-language-model, omni-modal-video-understanding, omni-model
Last push: 2025-03-28T00:56:46+00:00

## Health v2 (maintenance only)
Score: 29/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 13, release rhythm 35, longevity 53
- inputs: {"age_days": 753, "days_push": 524, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2534, forks 182 (observed 2026-08-28T04:06:59.101022+00:00)

## What it is
VITA is an open-source interactive omni multimodal large language model (VITA-1.5) that supports real-time vision and speech interaction, similar to GPT-4o. It provides training code, inference scripts, and demos for English and Chinese multimodal (image, video, audio) understanding and conversation.

## Use cases
- build a real-time voice and vision assistant like GPT-4o
- run an open-source omni multimodal LLM locally
- understand and chat about videos and images with speech input
- fine-tune a multimodal LLM on custom data
- evaluate a multimodal model on MLLM benchmarks like Video-MME
- demo real-time interactive AI conversation with camera and microphone

## When to choose
- you need an open-source GPT-4o-style real-time multimodal assistant
- you want to train or fine-tune an omni-modal LLM with vision and speech
- you need video understanding combined with voice interaction in English or Chinese

## When to avoid
- you only need text-only LLM inference
- you lack GPU resources for large multimodal models
- you need a production-ready commercial product with licensing guarantees (license is custom/unclear)

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, speech-recognition, computer-vision, chatbot
- domain: large-language-models, artificial-intelligence, computer-vision, speech-processing
- platform: python, cross-platform
- tags: multimodal, omni-modal, vision-language-model, real-time-interaction, video-understanding, speech-interaction, neurips-2025, gpt-4o-alternative, natural-language-processing, video, gpu, linux

## Member repositories
- VITA-MLLM/VITA (main) score 29

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:59.101022+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:25:15.025668+00:00, confidence not recorded.
  - readme: https://github.com/VITA-MLLM/VITA (fetched 2026-08-28T04:06:59.101022+00:00, sha e4354b745133)
- Data as of 2026-08-30T08:39:29.467469+00:00.