QMedia

QMedia is a free, open-source multimedia search engine for content creators. It searches text, images, and short videos, then combines their information in a multimodal retrieval and question-answering system. Its web service displays content and sources as cards with breakdowns of image, text, and short-video information. The project can be deployed locally for offline search and questions over private data. Its web interface, search and Q&A service, and image, text, and video model service can be deployed separately. The standalone model service provides an API for image and text encoding, video transcription, and image OCR. Documented model choices include CLIP, BGE, Qanything OCR, LLaVA, and Faster Whisper, with support for switching local Ollama models. The RAG service can use Google content search, but that option is configurable and disabled by default; video service use is also configurable and off by default. The RAG service depends on a separate mm_server service. QMedia is licensed under MIT and the project links to issue reporting and community groups.

QMedia brings multimodal search and Q&A together with options for local deployment and API use. Its service dependencies and configurable defaults are worth reviewing when planning a setup.

QMedia suits content creators who want to search across text, images, and short videos or run search and Q&A locally over private data. Its separate services and mm_server dependency are relevant for people setting up a deployment.

QMedia is an open-source search and question-answering system for creators handling text, images, and short videos. It is strongest for people who want to run the services themselves and need one workflow across those media; its modular deployment and local model options come with service dependencies to manage.

Overview

QMedia brings text, image, and short-video content into a single multimodal retrieval-augmented generation system for search and questions. Its web interface presents results as cards with sources and separate breakdowns of image, text, and video information. That makes it easier to inspect the material associated with a result, rather than relying on a match without context.

The project supports local deployment for offline search and Q&A over private data. Running it locally can keep the workflow centered on a user's own environment, but that alone does not establish a security certification or compliance status. QMedia is licensed under MIT, and its README points users to issue reporting, requests, Discord, and WeChat communities.

Key features

• Search across media: QMedia searches and extracts information from text, images, and short videos. This breadth suits collections where relevant context is split across formats; a tool built around just one media type may be a simpler fit for narrower needs.
• Multimodal Q&A: Its RAG system combines text, image, and short-video information for content search and question answering. The RAG service depends on a separate mm_server service, so it is not a single component to deploy and forget.
• Source-bearing content cards: The web service displays content and its sources in cards, with breakdowns of the image, text, and short-video information. That presentation helps users review what a result draws on.
• Configurable deployment: The web service, RAG search and Q&A service, and image/text/video model service can be deployed separately. That separation allows deployment choices, but also means an operator must account for the RAG service's mm_server dependency.
• Model options and API: The project documents local Ollama model switching and lists CLIP, BGE, Qanything OCR, LLaVA, and Faster Whisper among its model options. Its standalone multimodal model service can expose image and text encoding, video transcription, and image OCR through an API.
• Optional Google search and video service: Google content search is configurable and disabled by default. Video service use is also configurable and off by default in the RAG server configuration, so deployments that need those functions must account for their settings.

Pricing

QMedia costs free under the QMedia plan, with open-source software under the MIT License. There is no paid tier or quota stated for this plan, so it suits users prepared to deploy the software themselves rather than buyers seeking a packaged, priced service. The license makes the code available on open-source terms; it does not remove the work of operating the separate services or configuring optional functions.

Platforms

QMedia is available as an API, self-hosted software, and a web service. These are complementary ways to use the project: self-hosting supports local offline work over private data, while the standalone model service offers API access to encoding, transcription, and OCR. The web service presents results in cards, though the project also requires operators to manage service components for deployments using RAG.

Who it's for

QMedia is a good fit for content creators and technical teams with mixed text, image, and short-video collections who want search and Q&A in one system. It is particularly relevant when local operation over private data or direct use of the model service API matters. It is a less natural choice for someone who wants a fully managed workflow without responsibility for service dependencies and configuration.

Pros and cons

• Pro: One search and Q&A workflow spans text, images, and short videos. That can serve mixed-media collections without limiting the system to a single content type.
• Pro: Local deployment and separable services offer deployment flexibility. Local offline use can suit private-data workflows, and the services can be deployed independently.
• Pro: Model-service API covers encoding, transcription, and OCR. This gives API users access to several media-processing functions from the standalone service.
• Con: The RAG service relies on mm_server. Running multimodal RAG therefore requires operating a separate service.
• Con: Video service use is off by default. Deployments that need video functionality must change the configuration.
• Con: Local deployment is not a security or compliance assurance. Organizations with formal requirements still need to assess their own deployment.

Alternatives

For other options, see AI Video Search Tools.

• FrameQuery is worth considering for video-focused search where its free plan's metadata search, local/Drive/NAS sharing, and backup matter; its free tier has no monthly indexing hours, and face, voice, and AI-tag search require indexed video.
• DISKO may suit Mac and web users seeking screenwriting and shot-list tools alongside review links: its free cloud plan includes 2 GB, while its Pro plan is $12 a month at a founders rate.
• Twelve Labs is an API and web alternative for teams that prefer defined video-index limits: its free plan includes 600 minutes of video, 90-day index access, up to 10 hours and 100 videos per index, and five concurrent indexing tasks.
• ViSearch offers API, self-hosted, and web access with a free allowance of up to 20 minutes of video processing; its paid Basic plan charges $0.09, paid once, billed at $0.09/min.
• Vivu is an API and web option for users who want a stated monthly allowance: its free plan includes 20 indexing minutes and 50 search credits per month, plus a personal workspace and web app.
• Cinestar is a Linux and macOS alternative with a free MIT-licensed open-source core; local AI services require Ollama.
• Moment Search may fit Mac and Windows users who want on-device visual and speech search for a small footage library: its free Starter plan covers one footage source and up to 25 indexed videos, with clip export and clipboard copy.
• Scout Suite is a free, self-hosted, open-source multi-cloud security auditing tool, rather than a multimedia search alternative.

Verdict

Choose QMedia if you are a creator or technical team that needs to search and question a mixed collection of text, images, and short videos, especially when local operation or model-service API access matters. Its broad media scope and MIT license are the main draws; look elsewhere if you need a managed setup that avoids operating the RAG dependency and configuring video service use.