Agnes Video Generator

Troubleshooting & errors

HTTP 503 queue-full, 401 invalid token, failed or interrupted generations — and how to fix each.

Troubleshooting & errors

What are the system requirements?
Agnes Video Generator runs on Linux, macOS, and Windows. It requires Node.js 18+ and a modern web browser. Since all AI processing happens via cloud APIs, you don't need a high-end GPU — your local machine's hardware specs are not a bottleneck.
What API keys do I need?
Only the Agnes AI API key. Sign up at agnes-ai.com to get one for free. It covers text, image, video generation, and voice synthesis — all powered by Agnes AI's free models. No additional configuration needed.
Does it support batch video generation?
Yes. The Manuscript Video mode is designed for batch processing — it takes a long text, splits it into segments, and generates videos for each segment sequentially with unified narration and subtitles.
What happens if generation fails or is interrupted?
Agnes Video Generator has a checkpoint resume feature. If a generation task is interrupted (network issue, system restart, etc.), it automatically resumes from the last successful checkpoint when restarted, so you don't lose progress. Most failures are caused by transient factors such as model service fluctuations, network timeouts, or rate limiting — click Retry Task in the failure panel first, and the task resumes from the failed step (checkpoint-based), usually recovering on its own.
What should I do when generation fails?
Most failures are caused by transient factors such as model service fluctuations, network timeouts, or rate limiting. In the in-app failure panel, click Retry Task first — the task resumes from the failed step, and most cases recover on their own without resubmitting. If it still fails after several retries (≥ 2), the feedback area auto-expands, letting you copy the diagnostic info in one click and jump to a pre-filled GitHub Issue without describing your environment manually.
Why do I get an "HTTP 503" or "video queue is full" error from Agnes?
Agnes runs a shared free inference queue for video models such as Video 2.5 Flash, and it can stay saturated for 10+ minutes during peak hours. When it is full, every submission is rejected with HTTP 503 · video_queue_full and the raw message "video queue is full, please retry later (request id: …)". No task is created, nothing is consumed, and nothing is lost — this is Agnes-side capacity, not a problem with your prompt or setup. Since v7.0 the app no longer counts a queue-full rejection against the normal retry budget: it moves to a dedicated retry track (up to 15 attempts by default, tunable via AGNES_VIDEO_BUSY_RETRY_ATTEMPTS), the first retry fires immediately and later ones are spaced 30–60s apart. Since v7.1 the failure panel and the create-task toast also suggest the off-peak windows when upstream is historically smoother, shown in your own local time. To avoid Agnes 503 errors: retry during off-peak hours, switch to another video model (Video 2.0 uses a separate queue and often still accepts jobs), or add more API keys — each key has its own quota. One more note from the provider: the Agnes China site enforces real-name verification from 2026-10-26 12:00 (UTC+8) — unverified accounts get a lower API call rate — and upstream expects 503 / queue is full busy errors to improve significantly after that point.
Why do I get 401 / "invalid token" errors even though my API key looks correct?
A 401 Unauthorized or "invalid token" response usually means the API Key does not match the domain it is being sent to — for example, a key issued on the global site being used against the China-domestic endpoint api.agnes-ai.cn (or the reverse). Keys issued for different sites are rejected when sent to the wrong domain. As of v6.4.2, each key can be bound to its own access domain, and a one-click Auto-detect domains button probes each key across com / cn / cn_bak and fills in the matching domain automatically. In the API Key panel, pick the domain that matches your key, or just run auto-detect. Global-site keys should use apihub.agnes-ai.com, or the cn_bak fallback domain (apihub.agnes-ai.cn, which accepts both domestic and global keys).
How do I get help or report issues?
Feedback is available both in-app and on GitHub. When a task fails in the app, the failure panel offers Copy Diagnostic Info and Open a GitHub Issue buttons, automatically attaching the app version, task type, failed step, error message, and retry count to help maintainers pinpoint the problem. You can also visit the GitHub Issues page (github.com/lcy362/agnes-video-generator/issues) to check existing reports or open a new one.

Related Pages

Data & Report Basis

Answers about AI video capabilities, costs, trends, and trust on this page are supported by the following authoritative reports.

  1. Artificial Intelligence Index Report 2025Stanford HAI · Stanford AI Index 2025 (2025)

    Stanford AI Index 2025's authoritative data helps answer common questions about AI video generation capabilities, costs, and trends.

  2. Digital News Report 2025: AI-Generated Content and Trust in NewsReuters Institute, University of Oxford · Reuters Institute Digital News Report (2025)

    Reuters Institute 2025's survey data on AI content trust directly addresses user concerns about AI video quality and credibility.

  3. Global AI Video Generation Market Size & Share Analysis (2024-2030)Grand View Research · Grand View Research Market Report (2024)

    Grand View Research's AI video generation market sizing answers industry-outlook questions.

Why you can trust this

Written by a practicing developer and cross-checked against authoritative primary sources such as arXiv papers, vendor technical reports, and the Stanford AI Index.

S

SandGrid@lcy362

Author of Agnes Video Generator · Full-stack Developer

Independent developer and author of Agnes Video Generator, an open-source (MIT) AI video generation tool built on Agnes AI's free video models. Has helped 1,000+ creators produce AI videos at zero cost. Focused on making video generation models accessible and production-ready; content is based on hands-on practice and primary research.

View on GitHub

Last updated:2026-08-21

Ready to Start Creating?

"Making world-class AI belong to everyone." — Bruce Yang. It's completely free, no credit card, and you won't need a high-end GPU. Your first AI video starts at zero cost. Want to use Agnes AI's free video models? This is the easiest way in.

Clone the GitHub repo and launch in 2 minutes