Posts

Showing posts from November, 2026

Find Dead Air in AI Voiceovers With One FFmpeg Command

Image
Hi there. When I build a narrated video from AI voice clips, long stretches of silence can sneak in between sentences. My first assembly of a roughly 14-minute video had 109 gaps longer than 1.2 seconds. I did not hunt for them by ear. I measured them. Here is the command and the fix. What I saw I generated 166 sentences one at a time and joined them into a video of 859.7 seconds. A checker flagged 109 silent stretches of 1.2 seconds or more. Each sentence file carried a long quiet tail, and those tails added up. The detection command With FFmpeg installed, one line lists every silent stretch: ``` ffmpeg -i "video.mp4" -af silencedetect=noise=-50dB:d=1.2 -f null - ``` Each silence_start and silence_end pair in the output marks a gap. d=1.2 means only gaps of 1.2 seconds or longer are reported, so change that number to change the rule. The fix Trim the quiet edges of each sentence file but keep 0.18 seconds (180 ms) of breathing room on both sides. Compare the aud...

Running Two AI Jobs at Once: Give Each Its Own Folder

Image
Hi there. Running two AI jobs at the same time cuts the waiting. It works as long as the jobs do not share an output folder. I learned this when I rendered one long video and three short clips together. Here is what broke and the simple rule I follow now. What broke The video program I use deletes and rebuilds its working folder ( _work ) inside the output folder every time it starts. I pointed the long video and the shorts at the same output folder and started them together. Both commands began without an error. Their working files collided, so the results could not be trusted. The lesson A command that starts successfully does not prove the tool is safe to run twice at once. Before running two jobs in parallel, I check three things: Does the tool delete or overwrite its output folder on start? Do both jobs write temporary files with the same names? Is another job already running? If I cannot answer all three, I do not run them together. What I changed Every job gets its own...

Stuck .crdownload File? Test It Before You Delete It

Image
Hi there. A download stalls, and all that is left in your folder is a file ending in .crdownload . I hit this while saving an 8-second AI-generated video clip. Instead of deleting it or renaming it blindly, I tested the file. Here is the two-command test that saved the clip. What happened I generated a clip on an AI video site and asked for the 1080p version. The preview played fine and showed 8 seconds. The browser download then timed out, and only an .mp4.crdownload file remained. Why the file name is not the answer .crdownload is a temporary name the browser uses while a download is in progress. Deleting it because of the name can throw away a clip that is actually complete. Renaming it and using it can send a truncated clip into your project. The file's contents are what matter. The two-command test You need FFmpeg installed. Run these in Command Prompt: Read the basics: ffprobe -v error -show_entries stream=codec_name,width,height:format=duration -of default=nw=1 ...

Verify a Big File Upload With a Checksum: 3 Windows Steps

Image
Hi there. A "100% uploaded" bar is not proof that your file is safe. I archive large video bundles in cloud storage, and I stopped trusting the progress bar. Now I compare a checksum before and after the upload. It takes three steps and one built-in Windows command. What a checksum tells you A checksum is a short fingerprint calculated from the file's contents. Change one byte and the fingerprint changes completely. If the fingerprint of the original matches the fingerprint of the copy you download back, the copy is identical. Step 1: Fingerprint the original Open PowerShell and run this line with your own path: ``` Get-FileHash "C:\path\to\bundle.zip" -Algorithm SHA256 ``` Copy the long hash into a text file next to the file's name. Step 2: Upload, then download it back Upload the file, then download it again from the cloud into a different folder. Do not reuse the original folder, or you can end up comparing a file with itself. Step 3: Compare ...

Claude in Chrome Errors: 4 Browser Automation Snags, Fixed

Image
Hi there. You hand the browser to an AI agent, step away, and come back to a message that says the task stopped. If you run a one-person business, that stalled task is a delayed upload, a delayed report, or a delayed order. This post is a symptom-first guide for that moment. Each section starts with what you see on screen, then gives the response I use. There are four symptoms and three short habits behind them, and each habit has a label you can remember. 1. Start here: match the message to a response Read the first line of the error before you change anything. Each message points to a different response. "not in group" error: go to section 2 (Tab Check). "Page still loading" on a file upload: go to section 3 (Retry). Chrome window frozen while a conversion job runs: go to section 4 (One at a Time). A 429 error from a browser tool: go to section 5 (local work first). The diagram above carries three of these labels: Tab Check, Retry, and One at a Time. The fou...

Stock Photo Relevance: 3 Checks Before You Write Your Post

Image
Hi there. One of my free stock photos showed an airplane with an airline logo printed on its wing. The photo was free to use, and it still did not belong in my project. This post is a pre-publish photo check for solo bloggers who add free photos to their posts. It has 3 checks, one rule for switching to an AI picture, and a way to catch a bad photo before you build anything around it. The airplane, the mask and the toy rocket I picked photos for a video from Pexels. Three of them did not match their scenes. One showed a person wearing a mask. One showed a plane with an airline logo on its wing. One was a child's toy rocket, and it appeared in a short. Each one was wrong for its scene. A search term names a topic. It does not show you the picture. Open one frame before you build anything After that, I changed the order of my work. Before production starts, I open a representative frame and look at it with my own eyes. Do the same with your blog post. Pick the photo for your ma...

AI Image Tool Usage Limit Hit? 4 Steps to Keep Working

Image
Hi there. Picture a deadline a few hours away and the screen in front of you says something is wrong with your AI image tool. I run a one-person business, I am not a developer, and I make images with AI tools for my posts and videos. Over time I learned that the first job is not fixing the problem but naming it, because a usage limit, a 429 response, and a plain error banner each need a different reaction. This post is the routine I use to tell them apart and keep the work moving. First, read what the screen is really telling you Before I click anything, I sort the message into one of four kinds. A limit notice: the tool says you used up your allowance and often shows a time when it resets. A 429 response: the tool or the browser helper returns the code 429, which means too many requests were sent. An error banner: the chat panel says "failed", but the work itself may have been done. A single bad prompt: one request fails again and again while every other request works. ...

Scheduled Post Hits a Lock: Wait, Skip, Never Retry

Image
Hi there. I run a one-person online business, and two AI agents publish to the same blog for me. One of them has a scheduled job that posts at a fixed time while I am away from my desk. The question that matters most is what that job does when the other agent is already publishing. This post covers the three behaviors I rely on: wait, refuse a stranger's unlock, and stop instead of retrying. The scenario: a scheduled post meets a busy destination My scheduled job fires at its set time and wants to publish. At that moment the other agent may be in the middle of its own publish to the same blog. If both write at once, I can end up with a duplicate post or a half-updated page. I use a lock to prevent that, and the rule around it is short. Right before any outside change, the agent takes the lock for that blog. It releases the lock only after it has read the live result back and written the record. Making things stays parallel. Drafting, image work, review and local previews never...

Independent AI Review: Use a Fresh Reviewer, Not the Maker

Image
Hi there. You ask your AI to check the work it just finished, and it replies "All items pass." For a one-person business, that reply feels like permission to hit publish. I run my content with AI and I am not a developer, so I learned the hard way what that reply is actually worth. This post shows the habit I now run before I publish: one AI makes, a different AI reviews, and I fix only what fails. Start With the Checklist You Hand the Reviewer A fresh reviewer is only as good as the instructions it gets. A vague request such as "please review this" invites a vague answer. A list of specific checks gives the reviewer something to find. Here is the list I give now. Check every date and every duration by recalculating it. Check every number that came from an estimate and see whether the script still says it is an estimate. Check every pair of figures in the same sentence and confirm they were measured the same way. Check every claim of "first," "be...

Claude Code Usage Running Out Fast? One New Chat Per Task

Image
Hi there. When you run a one-person business on a flat monthly AI subscription, the usage allowance is your real budget. I'm not a developer, and I spent a month reading my own Claude Code usage records to see where that budget went. It went to one habit: keeping everything in a single, endless chat. Here is the three-step routine I use now: New Chat, Resume File, Continue. A flat-fee plan is a budget, not a buffet A subscription feels unlimited until the day you hit the cap in the middle of client work. So I stopped guessing and measured. Everything below comes from my own usage records, covering September 26 to October 25, 2026. These numbers describe my setup, so treat them as a pattern to look for in yours. 61 chats and 8,386 calls in total. Each call carried an average of 375K to 456K tokens of context. Nine chats with more than 300 calls each produced 62% of all cache reads. The ten biggest chats produced 67% of all new tokens. Two terms are worth knowing. Context is ev...