AI SHORT PLAY FIELD COURSE: ONE PERSON SHOOTS AN AI SHORT PLAY WITH A FULL PROCESS HINT AND SKILL PACK

Episode one, five minutes. The character will talk AI skitsI counted this step down by 27 dollars。

In the same amount, run at the official price of the volcano ark with the Seedance 2.5, about 1,360。

Let's start with the results

• A guy who doesn't have to buy a graphic card to make a short play that talks
• The video is about 2 gross, 5 minutes less than 30 cents in one episode; to be more refined and processed, 3 cents more than 1 $1 5
• From script to clip, every step has a direct copy of the hint, 8 paragraphs
• I FIXED THE H3 AND RELEASED IT. OPEN IT
• An amphibious comedy, a real person's short play

Here's the division of labour:

• Written, Mirror: Claude
• Art, character, scene, prop: GPT Image from ChatGPT, dream, Nano Banana, too
• Photography plus actor: Mini Max H3, running on RunningHub
• Editor: Clip
• You: Producer, story book, shot, board

A man is a show team. The order is just one sentence:Actors first, then action, then clips。

How to read this:

• First one: look at the "new hands route" below. Make it 10 seconds in an hour
• You want to do a whole set: press the plate one, two, three down
• I want to make it look better
• You want to get the tools: there's a light-scenes page at the end of the text, and there's a short play skill kit for Claude

First time, go this way: 10 seconds an hour

The last three plates are a whole episode. But for the first time, don't come up and do it for five minutes。

Let's start with a 10-second piece:One character, one scene, one line。 Once you run, you know what every step looks like, and you come back and do a whole set。

1. Write a line (5 minutes)

Use one-one-one-on-one script hints, but only one play, one character, one or two lines. For example, the sheep stood in the middle of the courtyard and said, "I'm actually a poet. I'm sorry

2. Two diagrams (15 minutes)

Using the trigger for plate one, 1.3, there are only two: one for the character (whole body, face, white bottom, shoes) and one for the scene (no one in the image). Identity plates, props, spectroscopes, for the first time。

3. Writing video alerts (5 minutes)

Claude is given the use of the lines and the two drawings, with the hint of plate two, 2.2, only one paragraph。

Run RunningHub (20 minutes)

PRESS THE 6-STEP OF PLATE TWO-THREE, RUN FIVE SECONDS PREVIEW, SATISFY AND OUT 10 SECONDS. A SECTION OF APPROXIMATELY 100 RHS CAN BE TESTED WITH A FREE DAILY SIGN-OFF; THREE CARDS ARE DRAWN AND NO MONEY IS PAID。

5. Clip with subtitles, export (10 minutes)

Import footage, identify subtitles, cross-reference lines, export 9:16。

After that, you ran the whole process。

And then you want to do a full set, and then you go down with three plates: spectroscopes, identification boards, reaction lenses, clippings, all of which are for later。

Board one: Pre-asset, pre-empt the script and "actress."

A lot of people came up to produce videos, and each part of the role was different and the story was not up to it。

The problem is not a video model, in the early stages. There were three things in the front period:The script, the mirror, the asset map.

It's only for words and pictures, one word for no money. When we get to the video, the change is a card. The problems that can be solved here are solved here。

1.1 Scripts: Only what can be captured in a camera

THE IA VIDEO MODEL CAN'T READ "HE'S UPSET" AND IT CAN ONLY READ "HE'S GOT HIS HEAD DOWN AND HIS FINGERS ON HIS HORN."。

SO GIVE AN AI A SCRIPT THAT'S DIFFERENT FROM A NOVEL THAT PEOPLE READ. REMEMBER THREE:

1. Psychological activity shall be translated into action, expression or inner monologue。

Instead of writing "judgment of the teacher" and "judgment of the teacher's eyes and his beard."。

2. Fixed format and sound type。

SCENE, ACTION, LINES, SOUND. IT'S ONLY IN THE SCRIPT THAT AI CAN TELL WHICH ONE IS THE PICTURE AND WHICH ONE IS GOING TO SAY。

3. One episode of five minutes around a major conflict。

First 3 seconds of conflict, ending on suspense. Based on approximately 2 seconds of every 10 words, a set of words 750 to 1,000 words is enough。

The sheep and sheep play is written like this:

The participants in this collection are sheep and sheep, judges, Lee Su-chil, Zhao Shusheng, and viewers

1-1 Courtyard of the ancient town courtyard [outside] [day]
The goats were in plain clothes, standing in the middle of the yard, in a state of shock。
Sneaky words, low laughter in the crowd。
Goat: Actually, I'm a poet。
♪ When the people blinked ♪ ♪ Then a moment started to laugh ♪。
Where's Lee Su-jae? Ignorant are fearless。
The judge of the teacher lifted his eyes and slowly covered his beard。
Judge the teacher (the monologue of the heart): No matter what poem he recites, he will never win if the husband is present。

Direct copyable 👇 (to Claude or DeepSeek)

You're a short playwriter, familiar with the rhythm on the tremor, the video number. Turn the next story into a short play of about five minutes。

Require:
1. The participants in this collection shall be listed at the beginning。
2. The title of each scene begins with the title of the scene: the name of the scene [internal/outside] [day/night], e.g.: 1-1 the courtyard of the ancient town court [outside] [day]。
3. The images and actions begin with a "snap" for each line and only write about what can be seen in the lens. Instead of writing "he wants to" his mind, the psychological activity is changed to action, expression, or monologue。
4. Line format: Role name (emotion/action) lines. The words in the heart are "the name of the role (the monologue of the mind):" and "the name of the role (the voice) is not in the image and the voice is still written in the image。
Important voices written in [sound]..
Total lines 750-1,000 words. There was a conflict within three seconds of the beginning, and the end stopped on the suspense and no early settlement。
Each sentence is a separate line, with only a script, without interpretation and analysis。

Story:
♪ Paste a piece of a novel or a story about you ♪

There's only one idea. No story? Let it write 3 300-word storylines, pick one, and then change it to a script。

1.2 Spectroscope: Step forward, starter can skip first

The script says "What's going on" and the mirror says "What does the audience see": how big, how long is the camera, who's in it。

A rookie can skip first, write a hint directly to plate two, and make a film. When you want to control the rhythm, open another spectroscope。

And remember four:

1. Switch to cut。

The answer is not to take one shot at the end, but to cut back and forth between those who speak and those who listen. That's how the sheep say:

1. Sheep says, "I'm actually a poet. Close view, 2 seconds
2. And then everyone laughs: Panorama, two seconds
3. Lee Soo-jae, "What's wrong with Shifu?" Ignorant are fearless. Close view, 2 seconds
4. Goats do not move: Pronouncement, 1 second

2. Emotions change。 Quiet as a shock, a shock as an anger, one shot at a time。

3. Three large moves。 PREPARATION, IMPLEMENTATION, RESULTS. ONE SHOT AND A WHOLE BUNCH OF MOVES, AI CAN EASILY MAKE A MESS。

4. Change of scenery and jump time, which must be cut。 Stay in the middle of one to two seconds。

Each shot is long, and the following hint is written, just remember two numbers: the line is about two seconds per 10 words, and the person who listens is about one second。

When the spectroscope is removed and two to four continuous lenses are synthesized into a "materials" (5-15 seconds), plate two is generated by "parts"。

Can be copied directly

You're a lens director with 20 years of experience. Take the following script out of the mirror。

The lens rule:
1. The switch is cut: the person who speaks, the person who listens, the person who speaks. Long lines must be cut to the listener。
2. Every change in character sentiment gives an independent shot。
3. Large action to prepare for execution results, complex action to remove 3-5 shots。
Change scenery and jump time shall be cut, and between the two scenes shall be accompanied by a one-two-second rotation lens。
5. Do not use the same view for three consecutive shots。

Length of time: Vision / Panorama 3-4 seconds, Midview 2-3 seconds, Proximity / close-up 2-3 seconds; line words per 10 words per 2 seconds; reaction lens around 1 second; orgasm 3-5 seconds; single lens not exceeding 10 seconds。

Output of a table, column: Mirror, length of time, time, space, and motion (fixed/push/push/screech/crack) Image content, lines and sound, Enter Role
"The content of the video" is written as a statement of the following: the painting, the environment, the location of the person, the action, the expression, the light, the image。

Following the table, the two-to-four continuous lenses are synthesized into a piece of material, 5-15 seconds each, indicating a correlation such as "Reflection Paragraph 1: Camera 1-3"。

PAINT STYLE (3D MANDARIN / 2D ANIMATION / REAL MAN, SELECT)
Playbook:
( paste here)

1.3 Assets chart: "Selection" followed by action

THE ASSET MAP IS YOUR ACTOR, SET AND PROP. H3 FEEDS NINE REFERENCE MAPS AT A TIME, AND THE CLEANER IT IS, THE MORE PEOPLE IN EACH SECTION LOOK LIKE THE SAME PERSON。

Five rules:

1. Full body, front, white, hands down, no things, no faces
2. YOU HAVE TO WRITE SHOES, NOT AI
3. It's about height, size, age, face, hair, clothing, and age
4. The adult role is eight
5. People don't have to design it alone

When the image was set, a role identification board was created: face, side, back, and a few faces and positions. When the video model gets it, it will recognize the man from all angles。

THE IMAGE OF THE PERSON, DESIGNED BY AI, SHOULD NOT BE USED AS A REFERENCE FOR REAL-PERSON PHOTOS, ESPECIALLY OF STARS: THE PLATFORM WILL STOP AND THERE WILL BE A RISK OF PORTRAIT RIGHTS。

Three rules:

1. No one in the picture
2. It's the same way you paint people
3. THERE WERE THREE CONVERSATIONS: A FIGHT, A FIGHT, A FIGHT. WE'LL JUST HAVE TO CHANGE IT ON THE TOP OF THE CHART

Props, two rules:

1. Pure white, no background, no one
2. Write the actual size, like, "The cup's about 8 centimeters in diameter, with one hand holding it." Otherwise, the video will be small

Direct copyable 👇 (first paragraph: send the script to Claude to allow it to list all assets and illustrations)

YOU'RE AN AI OUT OF THE PICTURE DESIGNER. ACCORDING TO THE SCRIPT BELOW, ALL PERSONS, SCENES, PROPS THAT NEED TO BE SHOWN ARE LISTED, AND THE ILLUSTRATIONS ARE WRITTEN SEPARATELY. JUST HINTS, DON'T FIGURE OUT。

People: Faces, whites, hands down, hands off, no face. Writing of height, size, age, face, eyes, hair, hair, clothing and shoes corresponds to the age of the story. The adult role is about eight heads, and the villain is not too ugly. People don't write。
Scene: No one in the picture. The scenes of the dialogue were written on each of the person's stations, one on each spot, one on the other, and one on the whole picture that would make it clear to all。
Props: Pure white bottom, no background and no characters, with actual sizes written (e.g. "about 8 cm in diameter, one hand holding")。

PAINTING STYLE (3D MANDARIN, UE5 RENDERING MASS/ 2D ANIMATION / REAL MAN, SELECT ONE). ALL CHARACTERS, SCENES AND PROPS REMAIN THE SAME。

Playbook:
( paste here)

It can be copied directly

See this figure, make a 16:9 role ID. White or white background, no environment, props, watermarks。

– A large frontal body on the center left; a space around the side, a back, a back, a back, a side, a position, and a three-faced face。
- All the drawings don't overlap. No face, no hands, no feet。
- All angles keep the same face, same hair, the same dress, the same body ratio。
– A simple character card in the corner: name, identity, character key。
– Clean, simple, like animated studio role-setting pages, not posters。

Can be copied directly 👇

Maintaining the image, light and set style of the scene, generating images of the past in the opposite direction: the left and right, the back and the back, the back and back, the back of the door and the outside, and a wall that corresponds to this architectural style. Nobody in the picture。

This is where it comes up: a script, a spectroscope (optional), an identification board for each lead, two or three images for each scene, and one key prop。

It took two minutes to get to the four questions:

• Does every lead have an identity card
• Is there anyone in the scene
• Did you write the size
• Did the total number of lines exceed 1,000 words

NEXT, FEED THEM TO H3, LET THE CHARACTERS MOVE AND TALK。

Board two: Video generation. Let the characters move and talk

The asset map is ready, and the next step is to turn it into a moving, talking clip。

I used Mini Max's H3, running on Running Hub. First of all, why not use the hottest Seedance 2.5 now。

2.1 WHY H3: CHEAP AND SOUND AT ONCE

Let's just say one thing:

• H3: Mini Max released video models on July 31, this year for a maximum of 15 seconds, a single generation of images and sound (for white, audio) and an open source on August 3。
• Running Hub: an online running ComfyUI workflow site. You can upload pictures, fill in hints, and you don't have to buy graphic cards or software。

To be honest, it's better to talk about the quality and complexity of the action。

But for a short play, a 5-minute episode requires dozens of pieces of material, and it's going to have to go over and over and over again, and it's worth a lot of money. I didn't think I'd do it

• RunningHub Professional Edition is now time-limited 99 yuan a month (159) for a multi-dollar package of 54,000 RH, 1 RH approximately 0.0018
• GENERATE A VIDEO OF ABOUT 100 RHS FOR 5-10 SECONDS, THAT'S.. One, two, maybe
• A 5-MINUTE SHORT PLAY, BASED ON 50 PARAGRAPHS (AVERAGE OF 6 SECONDS) AND THREE CARDS PER SEGMENT, WAS ABOUT 150 TIMES GENERATED, 15,000 RHSVideo this step, 27 more
• The same amount (5 minutes x 3 times = 900 seconds) is replaced by the model value of the volcano ark's official page (no reference video, 5 seconds, 7.56 dollars, approximately 1.51 dollars/s), which is around 1360 dollars; and the time-limited price for Running Hub 2.5, which is about 0.49 dollars/s for 8 months, which is about 440 dollars
• It's 99 bucks a month, and it's a three-and-a-half set, or 18 minutes or so

OF COURSE, A 2 GROSS IS A "USEABLE" SLOT, COMING OUT 640 P. IF YOU WANT TO BE MORE SOPHISTICATED, YOU CAN REPROCESS IT ON THE BASIS OF THE FIRST GENERATION. IN THE SAME 10 SECONDS, I MEASURED THREE:

• Save your money: JUST ONE, 640 P, ABOUT 100 RH, ONE 2 GROSS. PRACTICE, PLAY
• Fine: ONE + ULTRA-SPECTRUM AMPLIFICATION (AND FOLDING OF THE ORIGINAL), 1920 WIDE, APPROXIMATELY 185 RH, ONE PARAGRAPH 3 GROSS. IT'S FOR WORK
• Very good: TWO PICKS, A FINE SAMPLE OF THE WHOLE PART, 1920 WIDE, APPROXIMATELY 795 RH, ONE PIECE OF ABOUT $1 5. DELIVERY, CRITICAL LENS

[Approved: 185 and 795 contain a total price]

In my opinion, the processed effects are perfectly comparable with the Seedance 2.0. Even if one episode of the 50-selected footage is perfect, each one spends more than one piece, and one episode is 60 or 70, or less than one tenth of the official value of Seedance。

H3 Official API is charged in seconds: 768P about 0.5 yuan/s, 2K about 0.8 yuan/s, and 768P in 10 seconds is about $5. I'll take a piece of it. Three reasons: H3 is open-source, and my work is moving to the Quantified version of H3 (INT8) issued by RunningHub, with several accelerator nodes that run fast; Running Hub runs faster and cheaper at machine running hours; and I normally run 640 Ps, a little smaller than 768 Ps。

H3 AND TWO OTHER THINGS THAT MATTER TO ME:

1. Role consistency: A maximum of 9 reference maps can be fed at a time, and the role plates, scenes, and props can be included。
2. There are three ways to workReference video, graphic or front-end frame, digital person, daily film, replica, mouth type, 2.3, 2.4)。

2.2 Video alert: one word only

THE BIGGEST DIFFERENCE BETWEEN SHORT PLAY AND REGULAR AI SHORTS IS:The character has to talk, the play goes on and on。

So the hint is much more subtle than "one walking." The rules in the curriculum, I reduced them to six:

1. A paragraph should state only one thing。 5 to 15 seconds, only one action, one emotional turn or one key line。

2. Write position, then action。 Who's left, who's right, where's the door, who's got the props. If you don't write it down, the next one will move。

3. Follow the end of the paragraph。 The previous paragraph had stood up, and the paragraph began standing。

4. The lines are unchanged and the sound is divided into three words:

• And he said: "I am, indeed, a poet." I'm sorry
• In the heart: judge the teacher's monologue: "No matter what poem he recites, he will never win if he is present." I'm sorry
• People aren't in the picture, and the sound is like, "The sheep and sheep."

IF YOU DON'T WRITE SEPARATELY, AI WILL LET THE CHARACTERS OPEN THEIR MOUTHS AND READ THEIR HEARTS。

5. Do not stare at people who speak。 The line ends as long as it's long, as it's midway to people's reactions, and the lines keep going, and these seconds are replaced by "showing." If the cut is not good in a paragraph, the clip is added (3.2)。

6. Each reference chart is clear for use。 In my work stream, the hint box entered @ so you can choose the material: @photogram1 is the look of sheep and sheep, @photogram2 is the courtyard, which is better than throwing a brain in。

Can be copied directly to 👇 (sent to Claude to make the script or spectroscopy into H3-able.)

You're a short play director and an AI video alert engineer. Change the next script or spectrometer to a video alert that Mini Max H3 can use directly。

Rules:
For each period of 5-15 seconds, only one action, one emotional turn or one key line is dealt with. Different scenarios, at different times, have to be phased。
2. Each paragraph begins with the [position]: who is on the left, who is on the right, which side of the door, which side of the camera, and which key props are in hand and in which direction。
3. The action follows the end of the paragraph and does not repeat the action already completed in the previous paragraph。
4. The lines shall remain unchanged. The sound is written in three different ways: one says "" / "some inside" / "some in the outside" Each line is clear。
Reference material by purpose: @photogram 1 = profile of a given, @photogram 2 = scene, @photogram 3 = props. Written with sound reference @sound1 = sound of a given。
6. Write specific light and lenses: view, slot, fix or pull; fixed, non-push。
7. The sound will write only lines and the sound (footsteps, closing of doors) of the cause, without background music and without subtitles。
8. A fixed sound description (age, sound, speed) is written for each talking role, followed by a copy of the sentence as it is. A maximum of one or two lines should be inserted in the paragraph for 5 seconds。

Each paragraph output in this format:
[REFLECTION PARAGRAPH X] (X SECONDS)
[Reference]@phosphere 1:... @phosphere 2:..
[Stand]
[FIT] 0-X SECONDS: VISION, FLIGHT, WHAT'S GOING ON IN THE PICTURE, FACE AND MOVEMENT。
[Voice] Something says, "..." (Speaks) Sound:
[prohibited] don't aggravate people who don't exist in the book, don't show up with modern things, don't change hands, don't disappear, people don't change lines, don't subtitles and music。

Playbook or spectrometer:
( paste here)

2.3 Run RunningHub H3: From registration to download

I GOT MY H3 JOB OUT OF EXILE FOR THE FIRST TIME

Step 1: Registration。 Open runninghub.cn and sign in with a cell phone number. Every day you sign up for a free RH, try it out。

Step 2: Don't rush to buy membership。 First, you run the "new hands' route" with the money you sign and deliver, and you feel like you can buy the 99-dollar member. Members pay automatic monthly fees and do not want to remember to turn them off; the currency they send is valid and will not expire, so do not hoard。

STEP 3: OPEN MY H3 WORKFLOW。 Workstream Open Link. On the left, there are 9 picture frames, 3 audio frames and a " Disable Box " , with a hint box and parameters in the middle, with no movement on the right. The job was used at the open-source node ComfyUI-Mini MaxH3-Easy, thanks to the author。

Step 4: Upload reference maps and sound。 The number of the box is the number of @phosphere 1, @phosphere 2:

• Picture 1: The main character's identity card
• Picture 2: The scene of this section is only one
• Picture 3: Other roles, key props, or position panorama
• Audio 1: Role voice reference (2.4 speaking), not open without
• Use a few boxes to open a few in the Disable Box and keep the rest closed. By default, only pictures 1

Step 5: Punctuation with several parameters。

• Mode: Daily selection of "Reference Lives" ; by-censor selection of "Photo or First Curtain" (2.5 speaking); there are already well-defined lines, only mouth-to-mouth (2.4 speaking)
• Psych: When you put it in, redo the material with @ and make sure it's all right
• RESOLUTION: 640 P (SAVING; TRYING TO MAKE IT CLEARER)
• Image scale: change to 9:16 vertical screen (workstream default 16:9)
• Length: 5 seconds first filled

there is also a pit that is easy to step on: seed in the work stream is fixed. the fixed meaning is that the same diagram, the same hint, runs again almost the same. so if you want a new version of the "tick" card, you change the number, and if you want to fine-tune the tip, you keep it。

Step 6: Point run, download。 There are three running buttons in the top right corner, corresponding to the third set of machines, the faster the right, the more expensive. When you get out, you look at your mouth and look, and you download it. After downloading, you are named after the "paragraph number -- number two" like 03-2, and the clips are not random。

Two saving habits:

• Run 5 seconds preview, satisfactory 15 seconds。
• Multiple players try one by one。 Each character runs alone for a period of five seconds, confirming that it's long, and puts a few people together for a long shot. We'll run a few people together for 15 seconds, and we'll spend money once we fail。

2.4 Role voice: each section is the same person speaking

The greatest point of view of the short play is that the character speaks. But if the lamb is a bright juvie, the next one becomes a low-key old man, and the audience plays。

Three approaches, from simple to progressive:

1. Sound description of death, copying the same sentence in each paragraph。

Write a fixed sound description for each of the main players, such as "sheeps: a man's voice in his early 20s, bright, smooth, and careless." Each reminder is reproduced as it is, and not every time。

2. The first paragraph was satisfactory and the sound was cut out as a reference。

H3 CAN FEED UP TO 3 AUDIO REFERENCES AT A TIME, WRITTEN IN THE H3 ACCESS FILE, WHICH IS USED TO ALIGN THE SAME ROLE WITH THE SOUND IN DIFFERENT SEGMENTS. AFTER THE FIRST PARAGRAPH WAS SATISFACTORY, THE CHARACTER WAS CUT OUT FOR A FEW SECONDS (CLEAN, WITHOUT BACKGROUND), AND EACH OF THE SUBSEQUENT PARAGRAPHS WAS SENT IN, WITH THE HINT "@AUDIENCE1 IS THE VOICE OF SHEEP AND SHEEP"。

3. Sounding, then mouth-to-mouth with the "digital" model。

USE A PLACARD TO FIT THE LINE (EACH WITH THE SAME SOUND) AND THEN CUT THE MODE TO THE "DIGITAL PERSON" AND PASS ON AN AUDIO OF THE LINE. H3 WILL USE THIS VOICE TO GENERATE IMAGES AND MOUTHS THAT SPEAK, AND THE SOUND WON'T CHANGE. THIS MODEL CAN ONLY PASS ONE DRIVE AT A TIME。

Two other reminders:

• Don't put too many lines in a line。 THE DOCUMENT FOR H3 SAYS THAT A 5-SECOND CLIP IS A GOOD ONE OR TWO SENTENCE; IT'S TOO LONG, IT'S NOT CLEAR, IT'S TOO SHORT, IT'S TOO SHORT。
• Only with their own voice, or with an authorized voice。 Don't clone the stars, the voices of others。

2.5. Don't focus on smoking

I'M SURE THERE'S A LOT OF PEOPLE IN THE VIDEO: THEY'VE GOT A FINGER ON THEIR HANDS, THEY'VE GOT PROPS THAT SUDDENLY GET BIGGER, THEY'RE WALKING AROUND AND CHANGING THEIR FACES。

The order of treatment is as follows:

1. Cut if you can。 It's only a few frames at the beginning or at the end. Cut it off。
2. We can make it up。 IF THE PROBLEM IS IN THE MIDDLE, CUT DOWN THE FRAME BEFORE THE PROBLEM, RESIZE IT WITH A BIOGRAPHIC MODEL (E.G., RESIZE THE BIGGER CUP) AND THEN TAKE IT TO THE FRONT AND LET H3 FILL IT WITH THIS LITTLE BIT. THE FIRST FRAME MODEL OF H3 IS ALSO INCLUDED IN MY WORK STREAM: THE MODE CUTS TO THE "TURSON OR FIRST CURTAIN" AND ONLY OPENS A SINGLE PICTURE FRAME, FROM WHICH TO BEGIN; THE FIRST ONE IS THE FRAME, SPECIFYING THE BEGINNING AND END。
3. It's only the last time。 Before reticking, the hints or reference diagrams are reticked as they are, and the probability is still wrong。

Three other articles on prevention:

• A VIDEO IS GIVEN ONLY ONE SCENE. THE THREE SCENES WERE FED, AND AI DIDN'T KNOW WHICH ONE TO USE。
• The dialogue plays the "Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stand-Stamp-Stamp-Stamp-Stamp-Stamp-Stamp" and it helps to see where everyone is standing。
• The props are also written in the hints, and they're identical with the plate。

Clipboard three: Late, cut the clips into a set with a cut

The material came out, and the last step was to cut dozens of pieces together。

I cut it out, it's much simpler than da Vinci, and it's almost enough for a free version. It is easier to use a specialized version of the clipping on a computer。

Four things later:Sequencing, retweeting, sounding, subtitles.

3.1 Crude clippings: first sequence and then finish

1. In the script order, drag all the material into the time line. When a spectrometer is made, it is numbered by the texture paragraph。
2. Each paragraph is viewed from beginning to end. There's a problem with those frames at the end. Cut them. That's what the two and a half of the plates say, "Cut if you can."。
3. The little bit from the front frame is plugged back into the problem。

At the rhythm, the short play is faster than the movie:

• Start 3 seconds to conflict。 The first is the one that captures the most. The first part of the episode was "I'm actually a poet," followed by people laughing。
• Don't stop for too long without new information。 The gap between dialogue is also tightened。
• End on suspense。 Zhao Shusheng shouted, "He may have stolen this poem" and the people were murky and straight to the black field, so do not drag back。

3.2 White scenes: people whose voices continue to be heard

The board, speaking in the mirror, said that the white scene had to cut back and forth between those who spoke and those who listened. But what if the one that's generated is the one who's talking to the whole time

Add in clip:

1. Separate the sound of the speech (selects the clips in the clip, right-key "separated audio") and the sound remains in place and continues。
2. On top of that, you put on a "listen" reaction camera, like Lee Su-soo's eyes, the crowd's faces, covered in one to two seconds。
3. The audience saw: the first sentence looked at the sheep and said that the second sentence cut into the reactions of others and the sound was always the sheep。

Where's the reaction camera coming from? When you create it, you run a few more three to five seconds of reaction for each of the main players, not talking, just expression. In two squares, two hairs, 10 extras in one set, it costs about two dollars, but it's much better to cut。

3.3 Sound: 3 floors for white, sound, music

H3 HAS GENERATED WHITE AND SOUND EFFECTS, AND MUSIC HAS BEEN SAVED IN A SINGLE ONE IN THE CUT. IF EVERY PART OF AI GOES WITH MUSIC, IT'LL BE A LITTLE BIT OF A TUNE, SO THERE'S A "NO BACKGROUND MUSIC" IN THE CRICKET II TIP。

Three floors of sound. The volume is like this:

• Yes, maxThe audience must hear you
• Audio SecondFootsteps, closing doors. Follow the footage
• I'm the youngestKeep it down when someone talks

Two common problems:

• A few pieces of it。 Select all the clips to be evened at a time with the clips 'Uniform Sound'。
• The same role, individual voices。 The method of generation is described in plate II 2.4; if the clip is found to be leaking, the most economical remedy is to have only these strangulations, or to redo them。

Musical stylists or copyright-free music with a pre-commercial look at the authorization。

3.4 Subtitles: auto-identify and rephrase the script

AI OFTEN HAS THE WRONG WORDS AND THE WRONG CODE, SO TABLET TWO KEEPS H3 FROM SUBTITLES, SUBTITLES IN A CLIPPING RIGA。

1. "Recognition subtitles" with a clipping, which automatically translates into subtitles。
2. "Script Matching" to stick the lines in the script, it automatically corresponds to time, less wrong words。
3. WORD BY SCRIPT. THE LINES USED BY AI ARE SOMETIMES ONE OR TWO WORDS DIFFERENT FROM THE SCRIPT, AND THE SUBTITLES ARE CONSISTENT WITH THE SOUND: EITHER RETICK THIS PART OR CHANGE THE SUBTITLE TO THE ACTUAL ONE。

The style: put one third below the screen, get out of the way; one line is too long; one colour or brackets are changed in the heart, and the audience knows it。

3.5 Motion, Painting, Export

Motion: Dozens of pieces of material, the color would be a little different. Select the entire clip, add the same filter, or align the brightness and color temperature, and the whole set looks like a film。

Painting: MY WORK STREAM PICKED OUT 640 P. TO BE CLEAR, IT'S THE FINE AND EXTREMELY STYLISH PART OF THE PLATE, PROCESSED AT 1920 WIDTHS; THERE'S ALSO A GRAPHIC ENHANCEMENT IN THE CLIP, WHICH IS USUALLY FOR MEMBERS。

Export: STANDING BY: 9:16,1080P, 30 FRAMES, TREMORS, VIDEO NUMBERS, X IS ENOUGH。

Before we go, two last things:

1. Look at it from the top on the phone. Checklist of the penitentiaries of plate II 2.5: fingers, props, people with a change of face, doors with a change of direction。
2. MARK ON "AI GENERATION". ARTICLE X OF THE ARTIFICIAL INTELLECTUAL GENERATING SYNTHETIC CONTENT MARKING SCHEME, WHICH HAS BEEN IN OPERATION SINCE 1 SEPTEMBER 2025, REQUIRES USERS TO PUBLISH AI-GENERATED CONTENT, TO DECLARE IT ON THEIR OWN INITIATIVE AND TO MARK IT WITH THE MARKING FUNCTION PROVIDED BY THE PLATFORM. WHEN YOU HAVE A SHIVERING VOICE, VIDEO NUMBER, JUST HOOK UP WITH THE PLATFORM 'S AI GENERATION STATEMENT。

It can be copied directly

You're a clipper. The following is a script of this series and a good list of the materials I have produced (numbering, duration, content, problem). Help me out with a cut list。

Require:
1. Set the material in the sequence of events, indicating the number of seconds to the second and the end of each paragraph。
2. There are long lines that are always directed at the person who speaks, indicating which reaction lens, how long, and how loud the lines are。
3. List the lenses that need to be supplemented: what is missing, how long, who is responding。
4. Matching slots: Each paragraph writes emotions (stress / ease / invert) and starts, ends, and places with lines are "low"。
5. Finally, a separate list of all subtitles is given, the lines and scripts are identical, each with the word "speak" or "show " 。
6. There is a conflict at the beginning of three seconds and the end is on the suspense; if the order is not present, tell me how to adjust。

Playbook:
( paste here)

List of materials:
(Example: Paragraph 1, 8 seconds, sheep and sheep say, "I'm actually a poet," 7 seconds to deformed fingers)

Here, one short play is done。

Progress: Let's move from "seeable" to "good"

After the process went through, the next question must be: How do we make the picture more film-like

This is part of a photo and lens that I've been using to make films. The three most useful items are put here, and the quick check pages are attached at the end。

1. To be specific, not just "film."。

AI CAN'T READ "HIGH SENSE" AND IT CAN READ FOUR THINGS: WHERE DOES LIGHT COME FROM, HARD OR SOFT, COLD OR WARM, BRIGHT OR DARK. THE SIMPLEST TRICK: TO CHANGE THE LIGHT TO THE SIDELIGHT, THE FACE IS IN A STEREO。

2. A play locks a set of light。

In the same play, the same light is written in each paragraph: direction, warmness and the same time frame. The last one says "afternoon warmth" and the next one doesn't turn into "cold white light" or cut together like two films。

The sheep play, the day courtesies and the night candles, are two completely different sets of light: the day is bright, the afternoon is hard, warm and permeable; the night is when the red candle draws a contours from behind the side, and the dark is kept in the dark。

3. Six sets of one shot。

IT'S THE MIRROR, THE VIEW, THE ANGLE, THE SUBJECT, THE MOVEMENT, THE LIGHT. LESS, LIKE, AI GUESSED, WRONG GUESS WAS THE MONEY OF A CARD。

The first line is a template for your shot; the second line is a close-up of sheep and sheep hymns to the highest emotional point

Templates: Mirrors, (spectrum), (angular: who, what look), (motion: how much motion); (light: direction, soft, warm, dark); (focal section), film quality, real skin texture, dark stratums, clear facial shapes。

examples: fixed lenses, facial features, face-to-face, face-to-face, sheep and goats looking in the front, barely moving, with only their eyes glowing, with their chin; sidelight, warm and bright, light in their eyes and light in their eyes, with their background defaced; 50 mm, film tastes, real skin texture, with layers in the dark, with a clear-cut face。

FINALLY: THE NINE MOST COMMON MISTAKES COMMITTED BY AI

The pits of the short play are basically here in nine, and each one is followed by a section in which:

1. Each part of the role is different: A role identification board is produced and each part is fed (1.3)
2. It'll be the same for a while: one scene (2.5)
3. The people are moving, turning around: Each paragraph starts with a position and then with an action (2.2)
4. The part is open-minded: The line is divided into three words (2.2) of "inner monologue" and "information"
5. Same character. One voice at a time: sound description of death or reference to the first sound (2.4)
6. Long lines are dead: cover a human reaction camera (3.2)
7. It's the same as a suction card: workstream seed is fixed, reruns (2.3)
8. The props are too small: the actual size is written in the figure and in the hint (1.3)
9. There's no movie: Light-specific, one shot of all six

And finally, a piece of what I made out of this process

[ Embedded Video Post]

Appendix: Lightshot Speed Check Page and Claude Skills Package

Shadow and camera quick check。 I'M GOING THROUGH 25 PAGES WHEN IT'S STUCK: WHERE THE LIGHT COMES FROM, HOW BIG IT IS, WHETHER IT MOVES OR NOT. I MADE IT INTO A WEB PAGE THAT CAN BE READ ONLINE OR DOWNLOADED FROM PDF

i'm sorry, miriagent.com/light

AI SHORTPLAY SKILLS PACKAGE。 The process, I made three Claude skills: writing scripts, writing hints, H3 out. Put in Claude and say, "I'm going to do an IA short play, take me step by step," and it's going to take you to this. The download and installation methods are in the database

i'm sorry.com/library

The price of RunningHub membership and RH is what I saw in November 2026; the price of Feedance 2.5 on RunningHub is the time limit for August; and the prices of the volcano ark and Mini Max are at officially published standard prices. The platform will be priced, as you can see。

It's a long one, and it's suggested that the collection should be done in a consistent fashion。