MomentSearch

Paid APILinuxMacSelf-hostedWebWindows

What MomentSearch is

MomentSearch is open-source software for searching videos by what appears in them and what is said, then retrieving relevant moments. Users can upload videos or submit YouTube URLs for indexing. Each search checks both visual frames and transcript text and combines the results. Answers can point to clickable moments with keyframes, timestamps, synchronized transcript text, seen-or-said labels, and match scores. A confidence gate can abstain without calling an LLM when neither search branch clears its threshold. Without an LLM key, the software still returns ranked clickable moments but does not generate prose answers. Answer, embedding, reranking, and speech-to-text models can be selected by provider, with hosted and local options documented. The project packages Docker entrypoints and documents deployments to Fly, AWS, and Google Cloud. Its API does not authenticate requests: anyone able to reach its port can upload, ingest, ask, and delete data. The project recommends localhost binding or an authenticating proxy, along with private storage buckets. MomentSearch is single-user by default and released under the Apache 2.0 license.

Who it suits

MomentSearch suits people who need visual and transcript search across videos, with results linked to exact moments. Its documented Docker and cloud deployment options are aimed at users comfortable managing deployment and API access.

What it does well

Searches both visual frames and transcript text.
Answers can cite clickable moments with timestamps and keyframes.
Ranked moments remain available without an LLM key.
Models can be configured from hosted or local options.
Released under the Apache 2.0 license.

What to know first

The API has no authentication.
The application is single-user by default.
YouTube ingestion through SocialKit becomes paid after 20 free credits.
Cookie authentication cookies expire approximately every two to three weeks.

Limits

Without an LLM key, MomentSearch returns ranked clickable moments but does not generate prose answers. The documented SocialKit YouTube ingestion option has 20 free credits before it becomes paid; cookie-based authentication is free, but cookies expire approximately every two to three weeks.

Getting started

Open the project repository on GitHub.
Deploy the Docker image using the documented entrypoints and deployment guidance.
Configure storage and the documented database and retrieval services.
Set model provider options and keys for the answer and speech features you want.
Upload videos or provide YouTube URLs for indexing.
Bind the API to localhost or place it behind an authenticating proxy, and keep storage buckets private.

Verdict

MomentSearch combines visual and transcript retrieval and can link results to specific video moments. Its unauthenticated, single-user API calls for careful access controls before deployment.

Questions

Is MomentSearch free and open source?

It is listed as free and open source under the Apache 2.0 license.

Which platforms are listed?

The listed platforms are API, Linux, macOS, self-hosted, web, and Windows.

Can it answer without an LLM key?

It can return ranked clickable moments, but it does not generate prose answers without an LLM key.

What video sources can I index?

Users can upload videos or paste YouTube URLs for indexing. The documented SocialKit ingestion option provides 20 free credits before becoming paid.

Does the API require authentication?

No. Anyone who can reach its port can upload, ingest, ask questions, and delete data.

Which storage and infrastructure integrations are documented?

Storage options include local storage, AWS S3, GCP storage, GCP native storage, and Fly Tigris. The architecture uses Postgres, Qdrant, and Prefect Cloud.

Facts from MomentSearch's own pages

Abstention: A confidence gate can abstain without making an LLM call when neither retrieval branch clears its threshold. · github.com, 30 Sept 2026
Citations: Answers include clickable citations with matched keyframes, timestamps, transcript synchronization, evidence badges, and match scores. · github.com, 1 Oct 2026
Default models: The documented default stack uses OpenAI GPT-4o for answers, OpenAI text-embedding-3-small for transcripts, local CLIP for frames, fastembed for reranking, Whisper-1 for speech-to-text, and Gemini for speaker recognition. · github.com, 1 Oct 2026
Deployment: MomentSearch is packaged as a Docker image with api, worker, clip, and seed entrypoints and can deploy anywhere Docker runs. · github.com, 1 Oct 2026
Input: Users can upload videos or paste YouTube URLs for indexing. · github.com, 1 Oct 2026
Integrations: Documented infrastructure integrations include Postgres, Qdrant, Prefect, and object storage through Tigris, Amazon S3, or Google Cloud Storage. · github.com, 1 Oct 2026
License: MomentSearch is released under the Apache 2.0 license. · github.com, 1 Oct 2026
Maker privacy: Traversaal.ai says it uses strictly necessary authentication and usage cookies and does not use cookies for advertising, analytics, or tracking. · traversaal.ai, 30 Sept 2026
Model flexibility: Answer, embedding, reranking, and speech-to-text models are configurable by provider name, including hosted and local options. · github.com, 1 Oct 2026
Models: Answer, embedding, reranking, and speech-to-text models are pluggable by provider name, with OpenAI, Gemini, Anthropic, OpenRouter, xAI, Groq, Together, Fireworks, Mistral, NVIDIA, Azure OpenAI, Ollama, LM Studio, vLLM, Cohere, Voyage, Jina, and local CLIP options documented. · github.com, 30 Sept 2026
Purpose: MomentSearch lets users ask questions about videos and retrieves exact moments using what appears on screen and what is said in transcripts. · github.com, 1 Oct 2026
Retrieval: The system searches visual frames and transcripts in parallel, fuses the results, and reranks matching moments. · github.com, 1 Oct 2026
Retrieval-only mode: Without an LLM key, the application still returns ranked clickable moments but does not generate prose answers. · github.com, 1 Oct 2026
Security limit: The API is unauthenticated; anyone who can reach its port can mint upload URLs, ingest, ask, and delete, so the README recommends localhost binding or an authenticating proxy. · github.com, 30 Sept 2026
Security limitation: The API has no authentication, and anyone who can reach the port can upload, ingest, ask, and delete data. · github.com, 1 Oct 2026
Speaker recognition: Optional per-video speaker recognition uses Gemini video understanding to label who said each line. · github.com, 1 Oct 2026
Storage integrations: Object storage supports local storage, AWS S3, GCP storage, GCP native storage, and Fly Tigris, while the architecture uses Postgres, Qdrant, and Prefect Cloud. · github.com, 30 Sept 2026
Storage protection: The project recommends keeping buckets private and using presigned URLs for uploads and playback. · github.com, 1 Oct 2026
Uploads: Presigned uploads send gigabytes directly from the browser to object storage instead of through the API. · github.com, 30 Sept 2026
User model: The application is single-user by default and ignores X-User-Id and Authorization headers, routing every request through SINGLE_USER_ID. · github.com, 30 Sept 2026
YouTube ingestion limit: The documented SocialKit YouTube ingestion option provides 20 free credits and then becomes paid, while cookie authentication is free but cookies expire approximately every two to three weeks. · github.com, 30 Sept 2026

Where MomentSearch stands on Freedom251

AI Video Search Tools#17 of 24