DEV Community: Wenchao

Picking an AI video effect template vs writing a text-to-video prompt

Wenchao — Thu, 29 Oct 2026 08:00:12 +0000

When you want a short AI-generated video, there are two common starting points: a ready-made effect template, or a blank text-to-video prompt. I build SceneWish, so I think about this choice a lot. This post is my attempt to lay out when each one makes sense, without pretending either is always better.

What a template actually is

A video effect template is a pre-defined transformation. You supply an input, usually a photo, and the template decides the motion, style and pacing. Think of things like a hug effect, a hair-color swap, or a stylized animation applied to a portrait.

The appeal is constraint. Because someone has already tuned the effect, you don't have to describe camera movement, timing or style. You choose, upload, and wait.

The downside is the same constraint. You get what the template does. If your idea is slightly different from the available options, you can't easily bend it.

What a text-to-video prompt gives you

A prompt-only workflow starts from a blank box. You describe a scene in words and the model interprets it. That is flexible: you can ask for a fox walking through fog or a product on a rotating pedestal.

The cost is that you carry the design work. Vague prompts produce vague results, and getting a consistent outcome often takes several attempts. Different models also respond to wording differently, so a prompt that works in one place may need rewriting in another.

A rough way to decide

These are heuristics from building in this space, not benchmarks:

You have a specific photo and a known effect in mind. A template is usually faster. If you already have a picture and want it animated, an image-to-video workflow is a natural fit.
You need the same look repeated across many inputs. Templates are more predictable, because the effect definition stays fixed.
You have an original scene that doesn't exist as a template. Write a prompt.
You are exploring and don't mind iterating. Prompts give more room to experiment.
You are short on time and care about a repeatable result. Lean toward templates.

Writing a prompt that holds up

If you go the prompt route, a few habits help regardless of the tool:

Name one subject and one action. Prompts that try to stage three events in five seconds tend to blur.
Mention camera behavior explicitly (static, slow push-in, handheld) instead of hoping the model picks one.
State the setting and lighting in a short phrase.
Keep the clip's length in mind. Short clips rarely hold complex narratives.
Change one thing at a time between attempts so you can tell what caused a difference.

Using templates well

Template results depend heavily on the input image. Clear faces, decent lighting and a simple background generally behave better than crowded or heavily compressed photos. If a result looks off, try a different source image before concluding the effect itself is bad. You can browse the available video effect templates to see what kinds of transformations exist before deciding whether you need a custom prompt at all.

Combining both

The two aren't exclusive. A reasonable pattern is to start with a template to see if a ready-made effect covers your idea, and fall back to a written prompt when it doesn't. Another is to use a prompt to explore a concept, then look for a template if you find yourself repeating the same transformation.

Limits worth stating

Neither approach guarantees a usable result on the first try. Generated video can show artifacts, odd hands, drifting faces or unnatural motion, and the amount depends on the model and the input. Treat outputs as drafts and check them before sharing, especially if they involve real people. Also be mindful of consent when animating photos of other people.

Summary

Templates trade flexibility for speed and predictability. Prompts trade effort for freedom. Pick based on whether your idea already exists as an effect, how many times you need to repeat it, and how much iteration you are willing to do.

Disclosure: I'm the developer of SceneWish, the site linked in this post.

Daily Plain Game: Seven Estimation Rounds in Your Browser

Wenchao — Tue, 27 Oct 2026 03:20:44 +0000

Daily Plain Game (also called Plain Game, at plaingame.fun) is a free browser game built around estimation. Guess the scale. Climb the sky.

How it works

Each session has seven estimation rounds.
In every round you estimate a real-world quantity.
The closer your guess, the higher your plane climbs.
It is free, and there is no sign-up.

Practice anytime

Estimation is a skill you can sharpen. Open the game whenever you have a minute and see how high you can climb.

Play here: https://plaingame.fun/

Dessa AI Guide: An Independent Look at Block's Speech-to-Speech Model

Wenchao — Sun, 25 Oct 2026 15:41:32 +0000

Disclosure: I'm Chaozi, and this is an independent write-up. I'm not affiliated with Block.

Block has been publishing research on speech-to-speech models: systems that take spoken audio in and produce spoken audio out, without chaining separate speech-to-text, language model and text-to-speech stages. That direct approach can cut latency and keep tone, pacing and emphasis that a text pipeline throws away.

I put together Dessa AI Guide as an independent reference for people who want a plain-language look at this model. It covers what speech-to-speech means, why it matters for voice assistants and real-time conversation, and what to watch as the technology matures.

Why it matters

Lower latency: fewer hand-offs between components means faster replies.
More natural speech: prosody and emotion can carry through.
Simpler systems: one model instead of a three-stage pipeline.

If you're curious, visit Dessa AI Guide. It's an unofficial community resource, so please check Block's own announcements for authoritative details.

Feedback is welcome in the comments.

Two small projects I am building: Window Peeper AI and Plain Game

Wenchao — Sat, 24 Oct 2026 16:45:23 +0000

I'm building two small projects and sharing them here:

Window Peeper AI - an AI-powered tool.
Plain Game - a simple, no-frills browser game.

Feedback is welcome!

Building PokéBonk Wiki: an unofficial fan wiki for a free browser roguelike

Wenchao — Fri, 23 Oct 2026 00:36:38 +0000

I'm Chaozi, and I recently built PokéBonk Wiki, a fan-made, unofficial companion site for PokéBonk, a free browser 3D roguelike by Charcher. The game itself lives at https://pokebonk.com, and this project is not affiliated with it.

The wiki is published at pokebonk.info. I wanted something fast and simple, so the whole thing is static pages. There is no database and no server-side rendering at request time, which keeps hosting cheap and load times short.

The centerpiece is a 151-entry Pokédex reference. Entries use the main-series numbering, so anyone who knows the classic order can find a creature at the number they expect. Each entry is its own page with a consistent layout, which makes the reference easy to scan and easy to link to.

Beyond the reference, I wrote a set of guides that explain the game's ideas in plain language. I deliberately avoided publishing in-game numbers I can't verify, and kept the guides to general explanations of how a run is structured.

The site is available in four languages: English, Chinese, Japanese and Spanish. Each version has its own set of pages rather than a single page with a toggle, so every translation can be indexed and shared on its own.

Building a static multilingual site taught me a lot about keeping content consistent across languages. A shared structure for each page type helped most: I fill in the same fields every time.

If you enjoy the game and want a quick reference, feel free to take a look. Feedback on the structure and translations is welcome.

Party game night: 4 browser games that need no download

Wenchao — Thu, 22 Oct 2026 04:26:02 +0000

Disclosure: I'm Chaozi, an indie dev, and I built these four.

Noillirk: a multiplayer party game where you predict what most people answer. Good for mixed groups.
Farflung: a daily geography puzzle, good as a warm-up.
Daily Red: a daily Pokemon challenge against Red's fixed team.
2 Minute Drill: NFL last-two-minutes strategy for football fans.

All run in the browser. Tell me which one you would add to game night.

A quick Pokemon challenge format: beat Red one team at a time

Wenchao — Thu, 22 Oct 2026 04:25:56 +0000

Disclosure: I'm Chaozi, an indie dev, and I made the games linked here.

Red's team in the original games never changes, which makes it a good fixed puzzle. Daily Red turns that into a daily challenge against his fixed team.

If you prefer football, 2 Minute Drill puts you in the last two minutes of an NFL game.

For group play, try Noillirk, a party game about predicting the crowd, or Farflung, a daily geography puzzle.

Why a daily geography puzzle beats trivia for a 3-minute break

Wenchao — Thu, 22 Oct 2026 04:25:20 +0000

Disclosure: I'm Chaozi, an indie dev, and I built the games linked below.

Short daily puzzles work because they have a clear end. A geography puzzle gives you one question, a few guesses, and a result you can compare with friends. I made Farflung around that idea: one geography puzzle per day.

If you like the "guess what the crowd thinks" style instead, Noillirk is a multiplayer party game where you predict what most people answer. It works well as an icebreaker.

Other daily games I made:

Daily Red: a daily Pokemon challenge against Red's fixed team.
2 Minute Drill: NFL last-two-minutes strategy.

Feedback is welcome in the comments.

2 Minute Drill: building a browser football strategy game around the last two minutes

Wenchao — Thu, 22 Oct 2026 01:07:05 +0000

Disclosure: I built 2 Minute Drill.

2 Minute Drill is a browser American football strategy game based on the final two minutes of real NFL games.

Why the last two minutes?

That is where clock management, timeouts and play-calling decide games. Compressing the game to this window makes every decision matter.

Design notes

Real game situations give each scenario context fans recognise.
Strategy over twitch: you choose plays, not execute them.
Short sessions, so one scenario takes only a few minutes.

Feedback welcome: https://2mindrillgame.org/

Daily Red (REDLINE): a daily Pokemon challenge against one fixed opponent

Wenchao — Thu, 22 Oct 2026 01:06:58 +0000

Disclosure: I made Daily Red.

Daily Red (REDLINE) is a daily Pokemon puzzle. You build a team of six level-100 Pokemon and battle Red, whose team never changes.

The idea

Because the opponent is fixed, the game becomes a team-building puzzle: type coverage, speed tiers and move choice matter more than luck. Everyone faces the same challenge each day, so you can compare approaches.

Design notes

Level 100 for everyone removes grinding.
A fixed enemy team makes solutions shareable and debatable.
A daily format keeps it a five-minute habit rather than a time sink.

If you like Pokemon theorycrafting, give it a go and tell me what breaks: https://daily-red.app/

Building Noillirk: a party game about guessing what the crowd thinks

Wenchao — Thu, 22 Oct 2026 01:05:31 +0000

Disclosure: I'm the developer of Noillirk.

I built Noillirk because most party games reward knowing facts, and I wanted one that rewards knowing people.

How it works

Each round has 7 questions and you get 15 seconds for each. Instead of answering correctly, you answer the way you think most people will. Pick what the crowd picks and you score; pick the quirky answer and you don't.

Design notes

15 seconds is short enough that gut instinct beats overthinking, which is exactly the skill the game tests.
7 questions keeps a session around two minutes, so it fits between other things at a gathering.
Questions are everyday topics (food, habits, small talk), so nobody needs special knowledge.

What I learned building it

Tuning the timer was the hardest part: at 30s people debated, at 10s they panicked. 15s gave the most laughs. If you try it, I'd love feedback: https://noillirk.pro/

Building Farflung: a daily geography puzzle where rarer answers score more

Wenchao — Wed, 21 Oct 2026 16:59:55 +0000

I've been building a small browser game called Farflung and just put it online: https://farflunggame.fun/

It's a free English geography puzzle. Each round gives you 5 clues. For each clue you pick a country that fits, then drop a pin on the map.

The twist: several answers can be valid

Most clues fit more than one country. Any valid answer counts, but rarer correct answers score more, up to 500 "km" per round. So the obvious pick is safe, and the clever pick pays off.

Modes

Daily - one shared puzzle per day
Classic - 60 seconds per pin
Wander - untimed, relaxed
Unlimited - play as many rounds as you like

There are also leaderboards, an archive of past puzzles, friend challenges and leagues. No signup is required to play.

What I learned

Scoring by rarity makes players think beyond the first country that comes to mind, and it was more fun to tune than a plain right/wrong check.
Keeping the daily format short (5 clues) makes it easy to come back each day.

If you like geography games, give it a try and tell me what you think: https://farflunggame.fun/

DEV Community: Wenchao

Picking an AI video effect template vs writing a text-to-video prompt

What a template actually is

What a text-to-video prompt gives you

A rough way to decide

Writing a prompt that holds up

Using templates well

Combining both

Limits worth stating

Summary

Daily Plain Game: Seven Estimation Rounds in Your Browser

How it works

Practice anytime

Dessa AI Guide: An Independent Look at Block's Speech-to-Speech Model

Why it matters

Read more

Two small projects I am building: Window Peeper AI and Plain Game

Building PokéBonk Wiki: an unofficial fan wiki for a free browser roguelike

Party game night: 4 browser games that need no download

A quick Pokemon challenge format: beat Red one team at a time

Why a daily geography puzzle beats trivia for a 3-minute break

2 Minute Drill: building a browser football strategy game around the last two minutes

Why the last two minutes?

Design notes

Daily Red (REDLINE): a daily Pokemon challenge against one fixed opponent

The idea

Design notes

Building Noillirk: a party game about guessing what the crowd thinks

How it works

Design notes

What I learned building it

Building Farflung: a daily geography puzzle where rarer answers score more

The twist: several answers can be valid

Modes

What I learned