Developer adds free in-browser auto captions to YouTube Shorts converter using Whisper AI

·7 views

A developer has integrated automatic caption generation into Convert to Shorts, a free browser-based tool that reformats horizontal videos to the 9:16 YouTube Shorts aspect ratio without server uploads. The feature was the most-requested after the tool's launch, driven by the fact that most Shorts viewers watch without sound and manual captioning is time-consuming. To preserve the tool's privacy-first design, the developer ran OpenAI's Whisper Tiny speech recognition model entirely in the browser using Transformers.js and WebAssembly, avoiding any video upload to external servers. Audio is extracted and resampled to 16kHz mono using the Web Audio API before being passed to Whisper, which returns timestamped transcript segments. Captions are then burned into the video using ffmpeg.wasm with the ASS subtitle format, after simpler approaches like chained drawtext filters proved unreliable.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How AI Agents Help Engineers Recover Hidden Logic in SAP-to-Cloud Migrations

Migrating data from SAP to cloud platforms is technically straightforward, but preserving the business knowledge embedded in legacy systems — such as join logic, status mappings, and ownership — is far more complex. A synthetic case study involving over 5,000 SAP HANA calculation views, thousands of cloud tables, and parallel BI tools like Qlik and Spotfire illustrates the scale of this challenge. To address it, engineers have designed an Enterprise Data Discovery Assistant built on a bounded Snowflake Cortex Agent that queries four evidence layers: HANA repository code, BI artifacts, a reporting metadata catalog, and schema and lineage metadata. The assistant is read-only and does not modify source systems; instead, it surfaces recovered logic, ownership details, and dependencies to help engineers draft grounded SQL for validation. This approach reframes migration as an evidence-discovery problem rather than a pure code-conversion task.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

NeedFeed Launches Hyperlocal In-Kind Aid Platform to Replace Cash Charity

NeedFeed is a new mutual aid platform designed to connect neighbors through physical donations — such as food, clothing, and medical supplies — without involving cash transactions. The platform addresses three common failures in charity tech: cash diversion, involuntary public tagging of vulnerable people, and unverified generosity badges. Instead of tagging individuals in need, NeedFeed uses a steward system where a representative posts on a beneficiary's behalf only after recording verbal consent. Donations are verified through a two-party handoff process, meaning trust badges are earned through confirmed real-world exchanges rather than self-reporting. Built with Google Gemini AI and Snowflake integrations, the open-source project was submitted to the DEV Community Weekend Challenge: Generosity Edition.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer finds 93% of fully-signed Safe multisig transactions are permanently unexecutable

A developer building an automated executor for Safe multisig transactions ran a read-only detector against 1,299 Base mainnet Safes before writing any code, scanning blocks from Base's early history through block 50,776,046. The scan identified 366 fully-signed but unexecuted transactions, yet 339 of them turned out to be permanently dead — roughly 93% of the total. The root cause is that Safe executes nonces strictly in order, so if a different transaction consumes a given nonce, the original remains in the queue indefinitely with no indication it can never run. The naive one-line detector missed this entirely because it never compared queued nonces against the Safe's live on-chain nonce, a gap only exposed by an additional RPC call. This finding shifted the project's focus: instead of an execution engine, the core product became a structured refusal system that explains precisely why a transaction cannot or should not be submitted.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Web Tool to Remove Burned-In Text from Videos Using Alibaba API

A developer created a web interface called Video Text Remover to help remove burned-in text and subtitles from videos, inspired by a friend's e-commerce editing needs. The tool is built on top of Alibaba's VideoDetext API, which the developer found cost-effective and capable enough for basic use cases. Users can upload a video, detect or manually select the text region, and process the removal automatically. Testing showed the tool performs best when text appears over simple backgrounds, but struggles with moving objects or complex scenes. The developer is seeking community feedback on improving temporal consistency and handling more difficult inpainting scenarios.

0 comments Read more at DEV Community