Troubleshooting

First aid (try this first)

Before working through this page, try:

1. Close and reopen VidNexus. This restarts everything behind the scenes too.

2. Retry the failed action.

Many short-lived problems clear on a restart. If it fails again, come back here.

Render failures

Error ('ffmpeg' is not recognized)

ffmpeg (the video tool VidNexus uses) comes with the app. If it can't be found, the install is probably damaged.

Fix: reinstall VidNexus. There is no ffmpeg option to tick: it is always included. If it still happens, send a report from Settings > General > Report an issue.

Error (nvenc session limit)

NVIDIA cards for home use can only make a few videos at once. Other apps that record or stream can use up those slots.

Fix:

1. Close other apps using GPU encoding (OBS, Discord streaming, Chrome hardware acceleration).

2. Retry the render.

3. If it keeps happening, send a report from Settings > General > Report an issue.

Error (render hangs, no progress)

A common cause is a step that got stuck, often while making a motion graphic. If Stop build doesn't work within about 12 seconds, a Force stop button appears.

Fix: click Stop build, wait for the Force stop button, then click Force stop. The video is freed right away and you can start a new build at once. The stuck step stops at its next check without changing anything.

Error ('file is in use' on windows)

Windows sometimes holds on to a file for a moment after a video step ends. VidNexus tries again a few times. If you see this often, an antivirus scanner is usually holding the file. Windows Defender, Bitdefender and Norton are the usual suspects.

Fix: add %APPDATA%\VidNexus and %LOCALAPPDATA%\Programs\VidNexus to your antivirus's exclusion list.

AI model issues

Error (llm call failed / timeout)

Your AI model took too long to answer (the limit is 10 minutes per call) or returned an error.

Fix, in order:

1. Retry once. Most AI model errors go away on a second try.

2. Check Settings > Storage & Compute > Models. It shows whether Ollama is online.

3. If you use Ollama: run ollama list in a terminal to see if the model is installed. If not, run ollama pull <model>.

4. If you use a hosted provider (OpenAI, Anthropic): check your account for usage limits or billing problems.

Error (deep mode never finishes)

A Deep dive script asks the AI model up to 24 times, one after another. On a slow local model (for example a large model running on the processor), that can take 20 minutes or more. It's not stuck, just slow.

Fix: wait it out, switch to the default qwen3.5:9b or the larger qwen3.5:35b (both write quickly), or use a hosted provider for the Deep dive run.

Error ('no json in llm response')

The model answered in plain text when VidNexus needed a structured answer. Some models do this more than others, usually small models that aren't trained to follow instructions.

Fix: switch to a model that follows instructions better. Recommended local models: qwen3.5:9b (the default) or gemma4:12b. Cloud: any Claude / GPT-4-tier model.

Voice / ElevenLabs

Error (401 unauthorized)

Your ElevenLabs API key is missing or wrong. Paste it in Settings > API keys > ElevenLabs API key. Keys start with sk_.

Error (your ElevenLabs credit has run out)

You've used up your ElevenLabs monthly credit. The free tier is ~10 min of audio per month; the Starter tier is 30 min. Upgrade at elevenlabs.io/pricing or wait for the monthly reset. Then click Regen voice on the scenes that failed.

Symptom (voice sounds wrong / robotic)

The older ElevenLabs model (Monolingual v1) sounds flatter than Multilingual v2. Check Settings > API keys > ElevenLabs model. Leave it empty to use Multilingual v2. If one voice still sounds off, pick another narrator voice in the Voice tab.

Stock footage

Symptom (no clips found for a scene)

VidNexus searches Pexels, Pixabay and Coverr at the same time. If ALL three find nothing, your scene's search prompt is probably too abstract ("the ineffable nature of consciousness") or too niche ("19th century Prussian cavalry doctrine").

Fix: in the editor, click the scene, edit Visual prompt in the inspector, then click Fresh search. Use things you can see (people, places, objects). They work better than ideas.

Symptom (same clip appears in multiple scenes)

Clips are only kept unique within one scene. If two scenes have very similar visual prompts, they can get the same clip when it's the best match for both. Fix: click one of the scenes and pick a different clip under B-roll clips, or add more clips with + so several play in the scene.

Error (pexels 401 / pixabay 400)

Missing or expired API keys. Regenerate at pexels.com/api / pixabay.com/api/docs and paste them in Settings > API keys.

Research tools

Error (google trends 429 rate limited)

Google Trends has no official access for apps, and Google limits how often one internet connection can ask. Once flagged, they can block for 15 minutes to several hours.

What VidNexus does automatically:

• Asks Google one thing at a time, never several at once.

• Waits 10 minutes after Google says "too many requests", so new Topic pulse runs don't make the block longer.

• Saves good answers for 24 hours, so running Topic pulse again on the same topic is free.

What you can do: wait about 10 minutes for the pause to end, or use Keywords + Competitor for similar signal without touching Trends. If persistently blocked, your IP is on Google's longer-term list - try a VPN or a different network.

Error (youtube api quota exceeded)

The YouTube Data API gives you 10,000 units a day. Looking up channels or videos costs 1 unit; a search costs 100 units. A full Competitor run is ~3 units; a channel-history sync of 100 videos is ~5 units.

Fix: either wait until midnight Pacific time for the daily reset, or request a quota increase from Google Cloud Console. Increases are usually granted within 24 hours for legit use cases.

Error (pikzels thumbnail generation fails)

Pikzels is a paid service: each thumbnail uses a credit. The error says what went wrong, for example a missing or wrong API key, or a balance that is too low. Add credits at app.pikzels.com/platform, check the key in Settings > API keys, or clear the Pikzels API key there to stop using Pikzels.

Motion graphics

Symptom (live preview blank)

Switch to Final and back to Live in the top-right corner of the viewer. Sometimes the live preview doesn't start the first time, and switching once restarts it. If it keeps happening on several scenes, close and reopen the project.

Symptom (live preview shows old graphic)

The preview sometimes keeps showing an older copy of the video. Close and reopen the project. If it persists, restart VidNexus.

Symptom (auto-fill picked graphics you don't want)

Open Brand kits in the sidebar, open the channel's kit, and under Graphics you don't want, click each graphic to block. Auto-fill, the build and the Ctrl+K assistant won't add them on any video that uses the kit. For one video only, change the list in Video settings.

Channel history sync

Error (Add the channel's YouTube handle first)

Sync needs to know which YouTube channel is yours. Go to Channels > Edit on the channel and paste your handle in YouTube channel (an @handle, a UC... channel ID, or a youtube.com/@name link). Click Save, then click Sync videos.

Symptom ('Not matched to your channel's style' after sync)

Titles and Humanize say "Not matched to your channel's style · sync its YouTube history" when the selected channel has no synced videos. Make sure the context bar at the top of Research is set to the RIGHT channel (the one you just synced), not another channel with no history.

Symptom (channel history is stale)

VidNexus syncs each channel once a day while the app is open. The time on the sync button tells you how old the data is. To refresh now, click it to sync again. This updates view counts and pulls in new uploads.

Diagnostic bundle

Before you write to support, use Settings > General > Report an issue. Your license key and email are always included so support can reply. The optional diagnostic bundle adds your computer name and specs, service status, the names of your settings and recent crashes. It never includes API keys or your video content. Send it from the app, or use Open in email app or Copy. It works even when the license is blocked.

Step by step, with pictures: When something goes wrong.

What's next

Render pipeline: How the build works, step by step.

Community: Discord where you can ask for help live.