DEV Community: dongdong

16K Neural Image Super-Resolution and Batch E-Commerce Photo Restoration with Monipix

dongdong — Wed, 04 Nov 2026 07:05:37 +0000

Upscaling low-resolution product photos, vintage family portraits, or AI-generated concept art using traditional bicubic or Lanczos interpolation inevitably produces blurry edges, halo artifacts, and washed-out textures.

In this technical write-up, we share how we built Monipix — a browser-based AI image upscaler and creative workspace capable of 2x to 16x super-resolution (up to 4K, 8K, and 16K) — along with our HD Image Converter and open-source AI Image Upscaler Toolkit.

1. Tiled Neural Super-Resolution Without GPU OOM

Upscaling a 1024×1024 image by 8x or 16x produces a 16,384×16,384 canvas (268 megapixels). Running a single forward pass at that resolution exhausts VRAM immediately. To achieve artifact-free 16K output:

Overlapping Tile Decomposition: The input image is partitioned into overlapping tiles (e.g., 512×512 with a 64px feathering margin).
Degradation-Aware Restoration: Compression artifacts (JPEG blockiness, mosquito noise, motion blur) are removed prior to high-frequency texture synthesis.
Hann-Window Blending: Upscaled tiles are stitched using raised-cosine weights so tile boundaries are completely invisible even across smooth gradients.

interface UpscaleOptions {
  scaleFactor: 2 | 4 | 8 | 16;
  targetResolution: "4K" | "8K" | "16K";
  denoiseStrength: number; // 0.0 - 1.0
  faceEnhancement: boolean;
  outputFormat: "png" | "webp" | "jpg";
}

2. High-Frequency Detail Recovery for E-Commerce & Print

For e-commerce catalogs, anime illustrations, and print-ready posters, Monipix and the Monipix HD Image Converter preserve crisp typography, fabric weave, and facial details without over-smoothing.

Developers can also inspect our open-source benchmarks, scripts, and integration examples on GitHub at arkyu2077/ai-image-upscaler.

3. Links & Resources

Monipix 4K/8K/16K AI Image Upscaler: https://monipix.com
HD Image Converter: https://monipix.com/tools/hd-image-converter
GitHub Repository: https://github.com/arkyu2077/ai-image-upscaler

Automating Narrated Whiteboard Explainer Videos from Docs and Scripts with SpeedPainter & MCP

dongdong — Wed, 04 Nov 2026 07:05:07 +0000

Creating hand-drawn whiteboard explainer videos from technical documentation, product specs, or classroom notes usually requires hours of manual keyframing in desktop timeline editors.

In this technical guide, we walk through how we built SpeedPainter AI — a browser-based AI speed paint and whiteboard animation studio — and how developers and AI agents can automate narrated multi-scene explainer videos using our AI Whiteboard Explainer Video Maker and open-source SpeedPainter Explainer Video CLI & MCP Server.

1. Stroke-Order Extraction & Hand-Drawn Trajectory Synthesis

Unlike generic text-to-video diffusion models that hallucinate unreadable text or flickering diagrams, a true whiteboard speed-paint animation needs deterministic stroke progression:

Structural Edge & Contour Decomposition: Uploaded diagrams, illustrations, or generated storyboard frames are segmented into hierarchical vector contours (primary outlines first, internal shading and color fills second).
Continuous Path Ordering: We solve a locality-constrained path optimization problem so the virtual drawing hand moves naturally across the canvas without erratic jumps.
Synchronized Voiceover & Subtitle Alignment: In our AI Whiteboard Explainer Video Maker, each script paragraph is synthesized into natural narration while the stroke drawing duration is dynamically scaled to match the audio cadence.

interface SceneSpec {
  sceneIndex: number;
  narrationText: string;
  visualPrompt: string;
  sketchDurationSec: number;
  handStyle: "marker" | "pencil" | "chalk" | "none";
}

2. Automating Explainer Videos from Claude Desktop & Cursor via MCP

To let developers generate whiteboard animations directly from markdown READMEs or PR descriptions, we open-sourced SpeedPainterOrg/explainer-video.

You can wire it into Claude Desktop, Cursor, or Windsurf via MCP:

{
  "mcpServers": {
    "speedpainter": {
      "command": "npx",
      "args": ["-y", "@speedpainter/explainer-video-mcp"]
    }
  }
}

3. Try It Live

Web Studio (Image to Speed Paint Video): https://www.speedpainter.org/en
Script-to-Whiteboard Explainer Video Maker: https://www.speedpainter.org/en/explainer-video
GitHub Repository & MCP Integration: https://github.com/SpeedPainterOrg/explainer-video

Building a Portuguese (PT-BR) Audio & Video Transcription Pipeline with Whisper Large-v3, Pyannote Speaker Diarization, and MCP

dongdong — Wed, 04 Nov 2026 03:09:17 +0000

Transcribing multi-speaker audio and video in Brazilian Portuguese (PT-BR) comes with unique acoustic challenges: regional accents, colloquial expressions, overlapping speech in interviews, and low-bitrate voice notes (such as .opus / .ogg files from WhatsApp).

In this guide, we walk through how we architected Transcreva — a fast, browser-accessible AI transcription and subtitle generation platform with automatic speaker diarization — and how we exposed its workflow to AI coding agents using the Model Context Protocol (MCP).

1. Acoustic Transcription + Speaker Diarization Architecture

When you run standard speech-to-text on a 45-minute podcast or academic interview, a wall of unsegmented text is hard to read. To produce dialogue-ready transcripts with Speaker 1, Speaker 2 labels and word-level timestamps, we combine two stages:

ASR (Automatic Speech Recognition) with Whisper Large-v3: Generates high-accuracy Portuguese and multilingual text segments with word-level alignment.
Speaker Diarization with Pyannote: Computes speaker embeddings across sliding windows, clusters speaker turns, and aligns each Whisper word timestamp with the active speaker cluster.

interface TranscriptSegment {
  id: string;
  speaker: string; // e.g., "Speaker 1"
  startTime: number; // seconds
  endTime: number;
  text: string;
}

export function formatSrtTimestamp(seconds: number): string {
  const hrs = Math.floor(seconds / 3600);
  const mins = Math.floor((seconds % 3600) / 60);
  const secs = Math.floor(seconds % 60);
  const ms = Math.round((seconds % 1) * 1000);
  return `${String(hrs).padStart(2, "0")}:${String(mins).padStart(2, "0")}:${String(secs).padStart(2, "0")},${String(ms).padStart(3, "0")}`;
}

2. Handling WhatsApp Voice Notes (`.opus` / `.ogg`) and Video Subtitles (`.srt` / `.vtt`)

Two of the most common real-world workflows for Brazilian journalists, researchers, and creators are:

WhatsApp Audio Transcription: Voice notes exported from WhatsApp use the Opus codec inside an .ogg or .opus container. Our dedicated WhatsApp Audio Transcription workflow normalizes sample rates on the fly and outputs clean paragraphs or speaker-labeled dialogue.
Automatic Subtitle Generation: Video editors working on YouTube, Reels, or TikTok need frame-accurate captions. Using word-level timestamps from Whisper Large-v3, our Automatic Subtitle Generator segments lines by character-per-second (CPS) reading speed and exports ready-to-use .srt and .vtt files.

3. Integrating with Cursor & Claude Desktop via MCP (`transcreva-mcp`)

To bring transcription utilities directly into AI coding assistants (Cursor, Claude Desktop, Windsurf), we also built an open-source Model Context Protocol server: transcreva-mcp.

You can add it to your MCP configuration in seconds:

{
  "mcpServers": {
    "transcreva": {
      "command": "npx",
      "args": ["-y", "transcreva-mcp"]
    }
  }
}

Try It Out

Web App (Free Tier Available): https://transcreva.net/
GitHub Repository (MCP Server): https://github.com/arkyu2077/transcreva-mcp