Audio & video · checked 2026-10-19Can I vibecode BlitzReels?
KINDA· weekend project
The core is a pipeline you can run on your own machine: transcribe with Whisper, ask an LLM for the strongest moments, cut and crop to 9:16 with FFmpeg, and burn in word-timed captions. An agent gets that running in an afternoon, and open-source projects such as OpenShorts and AI-Youtube-Shorts-Generator show the shape. The weekend goes on what makes clips look finished: keeping a speaker framed as the camera cuts between people, caption styling that matches a brand, and a timeline editor to fix the picks. AI image and video generation, hosted rendering and the API are not worth rebuilding.
$39/moprice
$468you'd save a year
weekendbuild time
—people replaced it
What you need on hand.
DifficultyIntermediate · Some system setup or keys.
TimeAn afternoon for the first clips, a weekend to make them look finished
- Claude Code, Codex or CursorAny of them can run the prompt.
- Python 3.12mediapipe and faster-whisper pull in large packages, so allow a few GB of disk.
- FFmpeg and yt-dlpInstall them with your package manager; the prompt makes the tool check for them first.
- An LLM API key or a local modelUsed only to pick highlights from the transcript text.
Before you buildBlitzReels has no free plan, but OpenShorts is MIT-licensed and free to self-host. Try it before building, since it already covers the clip finder.
- 1. Install one agent: Claude Code, Codex, Cursor.
- 2. Make an empty folder and open it in the agent.
- 3. Paste the prompt below, press enter, and answer its questions as it builds.
The prompt
Build me a local long-video-to-shorts tool to replace BlitzReels. Requirements: - Python 3.12 CLI: `shorts make <file-or-url> --clips 5` writes ./out/<name>/clip-01.mp4 and so on. FFmpeg and yt-dlp run as external programs; fail early with a clear message when either is missing. - Transcribe with `faster-whisper` (WHISPER_MODEL in .env, default small) and keep word-level timestamps in ./out/<name>/transcript.json so a re-run skips this step. - Pick moments: send the transcript in chunks to an OpenAI-compatible endpoint from .env and ask for 20-90 second segments that start and end on sentence boundaries, each with a title and a one-line reason. Keep the best N and never let two clips overlap. - Reframe to 1080x1920 with FFmpeg: a face-tracked crop using `mediapipe` for one speaker, and a blurred-background fit when no face is found. Smooth the crop path so it does not jitter. - Captions: write an ASS file with `pysubs2` from the word timestamps, three or four words per line with the spoken word highlighted. Three styles live in ./styles/*.json (font, size, colors, position), chosen with --style, and are burned in by FFmpeg. - Pauses: `--trim-pauses` cuts silences longer than 0.7 seconds, found from word gaps. - Preview: `shorts serve` shows http://localhost:4830 with each clip in a player and buttons to move its start or end by 0.1 s and re-render only that clip. - Single user. No accounts, no telemetry, local-first: only the LLM request leaves the machine, and the README says it sends the transcript text. - Out of scope: AI image or video generation, stock B-roll, stacked two-speaker layouts, posting to social platforms, team features. - README: installing FFmpeg and yt-dlp, rough render times on CPU for a 60-minute video, and a reminder to process only videos you have the rights to.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
Then keep going
Next prompts, same folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own BlitzReels.
Add a two-speaker layout to the shorts tool in this folder. Requirements: - Detect faces with mediapipe on a frame every 0.5 seconds; when two faces are present for more than 2 seconds, render them stacked, each cropped to 1080x960. - Switch back to the single-speaker crop when only one face is visible, using a 0.4-second crossfade. - Place captions at the seam between the two halves. - A --layout auto|single|stacked flag; auto is the default. - Add a test with a short sample clip and update the README.
Add post text to the shorts tool in this folder. Requirements: - For each clip, ask the LLM for three hook lines under 60 characters, a 150-character description and five hashtags, based only on the clip's transcript. - Write them to clip-NN.json beside the video and show them in the preview page with a copy button. - Never claim anything the transcript does not say; include the transcript excerpt used. - A --lang flag for the language of the post text. - Update the README.
Add folder watching to the shorts tool in this folder. Requirements: - `shorts watch ~/Recordings` processes each new video file once it stops growing for 30 seconds, using the settings in a watch.toml file. - Keep a processed.json log by file hash so a restart never re-renders a file. - Run one job at a time and write errors to ./out/errors.log instead of stopping. - Send a desktop notification on macOS when a job finishes. - Update the README with a launchd example.
What you lose
- A browser timeline editor for adjusting cuts, captions and B-roll
- 50+ ready-made caption styles and a stock B-roll library
- Reusable brand Series and Story Kits
- AI image and video generation credits
- Hosted rendering, team seats, and the REST API, CLI and agent integrations
Why people still pay
Posting short clips is a chore that repeats: import, cut, caption, reframe, export. A hosted tool does it without installing FFmpeg or Whisper, renders in the cloud so a busy laptop stays free, and keeps captions styled the same each time. Teams add seats and the API on top.
BlitzReels or your own copy?
Over
Your usage
12h ×$/h
BlitzReels, 1 yr$468
Your own copy, 1 yr$40running + build session
You keep$428pays for itself in the first month
Estimate · Runs on your machine; Whisper and FFmpeg cost only time. Picking highlights costs a few cents of LLM tokens per hour of video. The build is token usage on your agent plan.
Take your data, then cancel.
1 · Take your data out of BlitzReels
- 1Export each finished clip from its project: the product renders finished videos, and paid plans add no watermark. Starter allows up to 4K output.
- 2The API lists a project's media and returns transcripts, using a Bearer key that starts with br_live_; the page we read doesn't explain how to create one.
- 3Caption styles, Series branding and Story Kits have no export in what we read: note the fonts, colors and positions you rely on.
- 4The terms contain no download or export clause, so don't assume the app keeps anything after you leave.
2 · Cancel BlitzReels
- 1Web subscribers cancel in BlitzReels billing settings; App Store subscribers cancel in their App Store account.
- 2Deleting your BlitzReels account does not cancel an Apple subscription.
- 3Cancelling stops renewal and access continues to the end of the billing period.
- 4The pricing page lists a 30-day money-back guarantee on paid plans; for web purchases the terms say to contact BlitzReels for a refund.
- The terms say content is deleted 30 days after the account is terminated, so download your finished clips and source files first.
- Trials need a card and convert to the chosen paid plan after 7 days, so cancel before day 7 if you are only testing.
- A failed or cancelled AI job isn't an automatic refund; send support the job ID for a usage review.
- Keep the same caption style and aspect ratios in your replacement so feeds don't change overnight.
The escape hatch
Don't feel like building it? These are already free.
OpenShortsOpen sourceMIT clip generator: AI moment detection, face tracking, subtitles and dubbing. Self-hosting is free with your own API keys and 8 GB or more of RAM; a hosted version starts at $12 a month.★ 6.8k · active Oct 2026 · docker to self-host
WebVisit ↗AI-Youtube-Shorts-GeneratorOpen sourceMIT command-line tool: LLM highlight detection, Whisper transcription and vertical cropping, with no watermark or credits.★ 5.5k · active Oct 2026 · one-click install
CLIVisit ↗FunClipOpen sourceMIT tool with a local Gradio page: speech recognition, subtitle generation and LLM-assisted clipping of long videos.★ 6.4k · active Oct 2026 · one-click install
WebVisit ↗
WebVisit ↗AI-Youtube-Shorts-GeneratorOpen sourceMIT command-line tool: LLM highlight detection, Whisper transcription and vertical cropping, with no watermark or credits.★ 5.5k · active Oct 2026 · one-click install
CLIVisit ↗FunClipOpen sourceMIT tool with a local Gradio page: speech recognition, subtitle generation and LLM-assisted clipping of long videos.★ 6.4k · active Oct 2026 · one-click install
WebVisit ↗
The numbers
BlitzReels pricing
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Starter | $39 | $29 | 30 exports and 15 AI credits a month, 10 projects, 25 GB storage, 1 seat, up to 4K output, captions, reframing, cleanup, AI image and video creation. No watermark. |
| Growth | $79 | $59 | 150 exports and 75 AI credits a month, 100 projects, 500 GB storage, 3 seats, Series with reusable branding and Story Kits. |
| Agency | $199 | $149 | Unlimited exports and projects, 250 AI credits a month, 2 TB storage, 10 seats, 5 concurrent exports. |
- Free tier
- The trial allows 1 import and 3 watermarked 720p exports; the paid plan you picked starts automatically after 7 days unless you cancel.
- Billing
- Monthly, or annual billed upfront at about 25% less. Plan changes are prorated to the next billing cycle. Cancel at any time and keep access to the end of the period. The page lists a 30-day money-back guarantee on paid plans. Payments go through Stripe. The page says launch pricing for early customers stays locked while the subscription remains active.
- Hidden costs
- AI credits (about $1 of AI work each) reset monthly and don't roll over; top-up packs that never expire cost $10 for 10 credits up to $100 for 150. Series and Story Kits need Growth or higher. The page gives no minute caps on exports.
- Source
- blitzreels.com
Also on the list
Related apps
No free plan. A 7-day trial needs a card and allows one import (720p, up to 15 minutes) and three 720p exports of up to 3 minutes each, with a watermark.
Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: A browser timeline editor for adjusting cuts, captions and B-roll, 50+ ready-made caption styles and a stock B-roll library.
About $39 a month on the Starter plan, $468 a year (checked 2026-10-19).
A browser timeline editor for adjusting cuts, captions and B-roll; 50+ ready-made caption styles and a stock B-roll library; Reusable brand Series and Story Kits; AI image and video generation credits; Hosted rendering, team seats, and the REST API, CLI and agent integrations. If any of those matter to you, keep paying.
Yes: OpenShorts (open source), AI-Youtube-Shorts-Generator (open source), FunClip (open source).