Build Log · linyan.io

One sentence in. Only one video came out.

The same single-sentence prompt, submitted to OpenArt Director and linyan.io. One returned a finished video in seconds. The other generated characters and scenes, then stalled.

July 2026 · Comparison

We tested both platforms' single sentence to video feature — the cornerstone of linyan.io's site.

The prompt

The test was simple: use the exact same single sentence as input to both platforms and compare the results. Our prompt was:

"ruth the woman and her mother in law naomi love each other and their relationship is approved by others"

Versions, on the record: linyan.io currently runs on Seedance 2.0. OpenArt Director orchestrates multiple third-party models under the hood; its run stalled before producing a final video, and the underlying video model for this session was not recorded. Both platforms on default settings, July 2026.

The results, side by side

linyan.io Completed
Watch on TikTok
Generated instantly — the finished video was on screen in seconds.
OpenArt Director Did not finish
OpenArt Director Ruth and Naomi result — characters and scenes generated, but no final video
Spent a substantial number of tokens, generated the characters and scenes, then got stuck — no final video despite the money spent.

Scorecard

linyan.ioOpenArt Director
Final videoDeliveredNot delivered
Time to resultSecondsStalled mid-pipeline
Partial outputCharacters and scenes only
CostOne generationTokens spent, no video
Verdict

Clear winner: linyan.io's single sentence to video feature works as advertised. OpenArt Director generated initial assets but did not complete the video.

Single run, same prompt, default settings on both platforms, July 2026. Engines and pipelines change fast — rerun this in six months and the result may differ.
Sakana AI
Field Notes
Scale Isn't Everything in AI

From the Stanford symposium: Why Sakana AI's lean approach may be the real disruptor in 2026's AI race.

July 2026
Same prompt, 30 seconds: where we lose to LTX
Build Log · linyan.io
Same prompt, 30 seconds: where we lose to LTX

One prompt, two engines, no retries. LTX-2.3 Pro rendered both characters. Our Seedance 2.0 pipeline never rendered the second one at all — and Pika's output suggests we're not the only ones.

July 2026
We used a fantasy feature film to break our own product
Build Log · linyan.io
We used a fantasy feature film to break our own product

70+ clips, two non-human characters, a dying firelight scene. What held up, what didn't, and what we're fixing next.

June 2026
AI F*** Me 3 Times Before Lunch
Field Notes
AI F*** Me 3 Times Before Lunch

A morning of mundane AI failures after rejecting dark AI spirituality. A true story from Yishun.

June 2026
The Jade Tithe
Case Study · AI Film
How we helped Sagelion make The Jade Tithe

A fantasy drama, an OpenClaw pipeline, and 70+ AI-generated cinematic clips. Here's how it came together.

June 2026
The future of agentic AI
Trends
The future of agentic AI — what changes in the next 6–12 months

From pilots to bounded autonomy: multi-agent orchestration, CLI agents, and why vertical beats “do everything.”

May 2026