videoctx
Advanced tools
Open, model-free video context layer for LLM agents: index once, query cheaply with any model.
Index a video once. Query it cheaply, forever, with any LLM.
videoctx ("video context") turns any video into an open, portable Video
Context index (.vctx, see SPEC.md) using only local, model-free
tools: ffmpeg keyframes, PySceneDetect shot boundaries, faster-whisper
transcripts, and optional one-line visual captions from any cheap VLM. An
agent answers most questions from the index's text layer for fractions of a
cent, and pulls individual frames on demand only when a question needs pixels.
videoctx index lecture.mp4 # one-time, local, free
videoctx show lecture.vctx # the text view the model reads
videoctx ask lecture.vctx "What diagram is shown when she explains backprop?"
Ask questions about videos the way you'd ask about a document:
Who this is for: developers and anyone comfortable running a few terminal commands. You'll need Python, ffmpeg, and an Anthropic API key for question-answering (indexing itself is free and offline). If that's not you yet, star/watch the repo: a no-install version is on the roadmap.
Native video ingestion re-tokenizes the whole video on every question: an hour of footage is ~250-350k tokens, every single time. Proprietary video platforms index into closed representations behind per-minute pricing. videoctx is the third option: an open on-disk format any model can consume, produced by free tools, where the marginal cost of a follow-up question on an hour-long video is about $0.02.
Full protocol and honest cost accounting (including prompt-cache write premiums) in paper/draft.md. Query model: Claude Opus 4.8.
Numbers below are single-touch runs of the frozen configs (v0.2: captions + OCR + evidence rules, Opus 4.8 loop; v0.3, 2026-10-21: adds text-evidence rules, verbatim transcript quotes and a final letter-vs-evidence check) under full cost accounting. Where we measured the same configuration more than once, we report the observed range: on n=150, single runs wobble by 2-3 questions; on n=30, by 2-3 questions (= 7-10 points). Treat sub-3-point deltas on this benchmark as noise.
Subset composition caveat: videos were taken in Video-MME dataset order, which clusters by domain. The medium and long cells are entirely Knowledge-domain videos; the short cells are 76-88% Knowledge (remainder Film & Television). Read these as Knowledge-dominated subsets of Video-MME, not the full benchmark; future splits will be domain-stratified.
Short videos, held-out test half (n=150):
| Method | Accuracy | $/question |
|---|---|---|
| videoctx v0.4 | 84.7% | $0.025 |
| videoctx v0.3 | 82.0% | $0.025 |
| videoctx v0.2 | 80.7% (equivalent configs ranged 80.7-85.3%) | $0.025 |
| Uniform frame sampling | 74.0% | $0.0172 |
| Gemini Flash-Lite native (pilot 30) | 70.0% | $0.0009 |
| Gemini Flash native (pilot 30) | 53.3% | $0.0047 |
Medium videos, 4-15 min (n=30, wide error bars):
| Method | Accuracy | $/question |
|---|---|---|
| Uniform frame sampling | 80.0% | $0.081 |
| videoctx v0.7 (contact sheet 60@512 in cached prefix) | 70.0% | $0.092 first-touch, ~$0.02 follow-up |
| videoctx v0.2 through v0.6 | 66.7% (near-identical failure set) | $0.051-0.062 |
| Gemini Flash-Lite native | 70.0% | $0.0034 |
Medium is our weakest cell and remains the named target: naive leads on raw accuracy. Ten questions failed identically across five configs (prompt rules, a text-view bug fix, dense captions, a 24-thumb contact sheet); the sixth config, a 60-thumbnail 512px contact sheet baked into the cached prefix, recovered one. The receipts tell a consistent story: these videos are fast-cut and fully captioned, and the remaining gap is visual detail bandwidth at question time, naive's one structural advantage. We publish the ladder rather than only its best rung.
Correction (2026-10-21): a text-view bug had been dropping the header line (duration, shot count) on every OCR-enriched index since OCR shipped; all v0.2/v0.3 numbers were measured under that bug. v0.4 is the corrected configuration (short +2.7 points, long back to 80.0%, medium unchanged).
Long videos, 30-60 minutes (n=30):
| Method | Accuracy | $/question | Follow-up $/question |
|---|---|---|---|
| videoctx v0.5 (dense captions) | 83.3% | $0.109 cold (measured $0.047, see note) | $0.021 |
| videoctx v0.4 | 80.0% | $0.099 | $0.019 |
| videoctx v0.2 / v0.3 | 80.0% / 76.7% (n=30 noise band) | $0.098 | $0.019 |
Cost correction (2026-10-22): the v0.5 long run started 57 minutes after the v0.4 run, and for the 7 of 10 videos whose text views were unchanged by dense captioning, it silently inherited v0.4's still-live 1-hour prompt caches, flattering first-question costs. $0.109/q is the cold-equivalent (first question $0.285, follow-ups $0.021 as measured). Benchmark runs are now cache-salted per config so this cannot recur. | Uniform frame sampling (60-frame cap) | 70.0% | $0.133 | $0.133 | | Gemini Flash-Lite native | 76.7% | $0.022 | $0.022 (5-9 min latency/q) | | Native ingestion (published Pro rates, est.) | pending | ~$0.50 | ~$0.50 |
One-time index cost: ~$0.10-0.25 per hour of video (captions + local OCR + local embeddings), pennies otherwise.
Cost correction history: early short-video figures ($0.0067/q) undercounted prompt-cache write premiums; the metering was fixed 2026-10-18 and all figures above are fully loaded. Details in the paper's Section 7.3.
git clone <repo> && cd videoctx
uv venv && uv pip install -e .
Requires ffmpeg on PATH. videoctx index and the MCP server's core tools
run fully locally with no API key. videoctx ask uses the Anthropic API
(ANTHROPIC_API_KEY or an ant auth login profile).
claude mcp add videoctx -- $(pwd)/.venv/bin/videoctx-mcp
The calling agent's own model reads the text layer and frames directly; the
server makes no model calls and needs no key. Tools: video_overview (cheap
text layer), survey_video (one thumbnail per shot), get_frames(path, t0, t1, detail), transcript_search, find_moments (semantic visual search,
needs videoctx embed once per index), index_video. The server's instructions
carry the evidence rule: visual answers must cite fetched frames, never
general knowledge.
uv pip install -e ".[eval]"
python eval/prepare_videomme.py --duration short --videos 100 --out eval/data
python eval/split_dev_test.py eval/data/videomme_short.jsonl
python eval/run.py eval/data/videomme_short_test.jsonl --methods videoctx,naive
eval/methods.py documents each baseline. Cost accounting counts cache
writes at their real premium (1.25x / 2x by TTL); see the paper's Section 7.3
for the accounting traps this avoids.
The .vctx directory is plain JSON + JPEG: manifest.json (provenance,
parameters, source hash), index.json (shots, keyframes, transcript,
captions), frames/. Consumers must ignore unknown keys; enrichment layers
(OCR, embeddings) are reserved in SPEC.md v0.1.
Working and benchmarked: indexing, query loop, MCP server, eval harness. Next: medium-duration subset, Pro-tier native baseline, cross-model portability runs (same indexes via non-Claude models), OCR and per-shot embedding layers, spec v1.0.
Apache-2.0. Copyright 2026 Fred Dunyo.
FAQs
Open, model-free video context layer for LLM agents: index once, query cheaply with any model.
We found that videoctx demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Security News
The PHP Foundation says Chainguard broke a security embargo ahead of critical RCE patches, as leaks disrupted the coordinated release.
Security News
PHP releases emergency patches for two critical RCE vulnerabilities as agencies urge organizations to take unpatched sites offline.
Threat Research
/Security News
A supply chain attack compromised LiteLLM 1.107.0 on PyPI and Docker, spreading malware that steals cloud credentials, developer secrets, and AI provider keys.