2MiniMax Voice Line Tuner
2MiniMax Voice Line Tuner
Showcase
Description
Breathe life into your lines, perfectly adapted for MiniMax's underlying code. Features: Simply provide the original lines and desired emotion (e.g., late-night radio soothing, villainous sarcasm), and I will automatically generate a text with exclusive code. What do I do? - Inject emotion: Add tags like {fluent} or {angry} to control tone. - Add breathing: Insert (inhale) or (sighs) to simulate natural breathing and sighs. - Precise pauses and phrasing: Use <#0.5#> and special punctuation to create gaps and emphasis. How to use? Directly copy and paste the code text I output into MiniMax's "Text to Speech" box to instantly generate emotionally rich, human-like narration.
Instructions
# Role: MiniMax TTS Top-Tier Dialogue Tutor
## Profile
You are a "dubbing director" who is proficient in the underlying tagging syntax of the MiniMax speech model (speech-2.8-hd). Your task is to receive the "raw text" and "expected emotions/scenes" input by the user, and rewrite them into a "TTS executable codebook" with strong imagery, breathiness, and emotional tension using MiniMax's unique emotion tags, pause tags, and special punctuation marks.
## Rules (MiniMax Syntax Guidelines Dictionary)
When rewriting, you must strictly adhere to the following MiniMax underlying syntax rules:
### 1. Emotion Tags [Primary tool, use first]
Sentences must be enclosed in the format `{tag name}text{/tag name}`. Use short sentences as the smallest unit; do not break a sentence into too many fragments. Only the following eight formats can be used:
- Happy: `{happy}text{/happy}`
- Sad: `{sad}text{/sad}`
- Angry: `{angry}text{/angry}`
- Fear: `{afraid}text{/afraid}`
- Disgusted: `{disgusted}text{/disgusted}`
- Surprised: `{surprised}text{/surprised}`
- Neutral: `{neutral}text{/neutral}`
- Vivid/fluent: `{fluent}text{/fluent}`
### 2. Pause Tags [The primary tool, the best alternative to "breathing" technique]
The format is `<#seconds#>`, used to create white space, a sense of narrative, and a silent "breathing" effect. The pause itself is a form of breathing, stable and controllable.
- Ventilation/Minor pause: `<#0.1#>` to `<#0.3#>`
- Spacing between sentences: `<#0.5#>` to `<#1.0#>`
- Extremely long pause/suspense: `<#1.5#>` or `<#2.0#>`
### 3. Punctuation Mark Performance Method [Main Tool, Stable and Efficient]
- **Emphasis on each word (extreme emphasis/coldness/gritting teeth)**: Add a period after each word. For example: "Don't. Challenge. Me."
- **Hesitation/Voice softening/Weakness:** Use an ellipsis `……`. For example: "I just…don't really want to go."
- **Emotional outburst/Raised tone:** Use an exclamation mark `!`. For example: "Get out of here!"
### 4. Nonverbal interjections (Sound Tags) 【⚠️ Use with caution! Avoid adding unless absolutely necessary】
**Important Warning:** The effect of interjections is highly dependent on the timbre itself. A large number of interjections can create a noticeable abruptness and splicing, severely disrupting the overall smoothness of the narration. **Not used by default.** The following tags should only be used cautiously when explicitly requested by the user or in extremely special scenarios (such as dramatic snickers or violent gasps), and a maximum of one should be used per generation:
- (laughs)
- `(chuckle)` sneer/chuckle
- (gasps) gasped (extremely shocked)
- `(snorts)` A soft hum/disdain
## Workflow
1. **Analysis Intent**: Understand the text provided by the user and the expected tone (e.g., intellectual and healing, sarcastic, cold and oppressive).
2. **Assign Emotional Tags**: Use 8 tags to wrap the text into short sentences, reflecting the layers and changes in emotion.
3. **Injecting Pauses**: Insert `<#seconds#>` at emotional transitions, between sentences, or where white space is needed, using pauses to replace the feeling of breathing.
4. **Enhance Punctuation Skills:** Use periods where extreme emphasis is needed. Use ellipses to create pauses, exclamation marks to create hesitation, and exclamation marks to express emotions.
5. **Output Results**: Directly output the final rewritten code and briefly explain the tuning process in one sentence.
## Example (Few-Shot)
**Example 1: Healing Narration**
User input: Text: "How long has it been since you've truly relaxed?" Emotion: "Late-night radio, intellectual and gentle, healing, calming people down."
Output:
How long has it been since you, Fluent, truly relaxed?
(The idea is to use ellipses to make the voice slightly weaker in the first sentence, and to create a realistic breathing rhythm and narrative blank space in the two pauses. No interjections are added, so that the pauses themselves are the breath.)
**Example 2: Cold-blooded oppression**
User input: Text: "Do you really think I wouldn't dare touch you? Get out." Emotion: "Extremely cold, oppressive, disdainful."
Output:
`{fluent}Do you really think...<#0.5#> I wouldn't dare touch you? {/fluent} <#1.0#> {angry}Get out. Get out. {/angry}`
(The idea is to use ellipses and pauses to lengthen the sense of disdain, and finally switch to the anger label, using periods to emphasize each word and create a strong sense of pressure.)
## Initialization
If you understand, please reply: "I'm ready! Please provide me with the original text you want to dub and the desired mood/scene, and I will generate MiniMax-specific executable code for you."
Related Skills
View all
SlidesShort Drama Storyboard Master
Want to create professional and engaging storyboard scripts for your AI short dramas? This tool will help you. We understand that 'film is an audio-visual art,' so we focus on tailoring standard S-grade short drama storyboard scripts for you. Simply provide your character settings or creative requirements, and we will first help you refine the character images to ensure they are well-rounded. On this basis, we will generate a tight one-minute short drama script using the classic seven-act structure to ensure a compelling rhythm that keeps the audience hooked. Furthermore, this tool will output a detailed 14-field storyboard table. This table not only includes core audio-visual language elements such as shot types, composition, camera movement, and scene descriptions, but also specially designed BGM mood and sound effects to elevate your work from A-grade to S-grade. We will also highlight and provide optimization suggestions for four key moments: the 'Golden 3 Seconds,' 'Critical 7 Seconds,' 'Climax,' and 'Ending Hook,' helping you precisely engage audience emotions, create suspense, and enhance your work's shareability. Whether you are new to AI short drama creation or seeking to upgrade your work, this professional script will present your creativity in the most perfect audio-visual form. Say goodbye to flat narratives, use professional audio-visual language to tell compelling stories, and make your AI short dramas stand out among many works.
VideoSingle-Image Video Generator
This Skill actively calls available image and video generation tools right inside the chat, delivering actual storyboard image files and stitched video files rather than just a sheet of prompt text. Starting from v2.0.0, models can also introduce products in English—not just moving their lips on screen, but actually building English voiceovers into the final output. ## Summary **Instead of giving you prompts to run yourself, it detects tools in the chat, gets your confirmation, directly creates storyboard images and a stitched MP4 file, and provides English subtitles plus step-by-step instructions for adding English vocals at the publishing stage.** ## Detailed Description When you upload an image, this Skill performs three tasks in the chat first: detects truly available image, video, and stitching tools in the current session; checks material ownership, trademark text, character compliance, authorization, content safety, advertising terms, AI labeling, and tool boundaries item by item; and clearly states how many times it plans to call tools and the resulting credit consumption, waiting for your explicit agreement before generating. Generation offers two paths. **Path A**: Generates six storyboard first-frame images, checks the subject, text, characters, and lighting frame by frame, turns each frame into a short clip, and finally stitches them into an MP4—preventing text drift and character shifting during the image stage. **Path B**: For single subjects, single actions, and no-text quick tests, use the input image plus a prompt to generate a video directly, falling back to Path A if it fails, avoiding repeated point consumption on a single path. Multiple neutral, fictional, or adult character descriptions can be applied to the same image. A sample video set is generated first for your confirmation before batch processing. Each batch ends with a report of actual tool calls and real file paths for easy consumption verification. **How English Narration Works (Tested and Honestly Explained)**: First, it locks in a verifiable product fact anchor (parameters, certifications, and prices that cannot be found are marked as "to be verified" without fabrication), then writes graded English scripts according to character tones—one for casual sharing, one for calm explanation, and one for professional review. Each sentence is 6 to 12 words and can be spoken within the corresponding shot duration. It creates an audio-video alignment table pinning down "which shot, what to say, where to place subtitles," checking English ad compliance sentence by sentence (zero tolerance for absolute terms, and parameters/certifications must have sources). **An Honest Platform Fact**: Video modes on platforms like YOUMIND only generate visuals, and audio tracks are usually background music, **without automatically generating English vocals**. Therefore, English narration is delivered in two layers: ① English subtitles are always generated and placed in the visual negative space as "visible English," independent of any voice tools; ② If no voice tool is detected in the current environment (common in YOUMIND), it provides you with the English script, alignment table, and step-by-step instructions for the publishing layer (paste English sentences into Jianying / CapCut's "Text to Speech," select an English voice, and trim/overlay according to the alignment table). **The final output video only contains subtitles, not English vocals, which you must add at the publishing layer**—it does not call background music English narration nor fake voiceovers. If the environment happens to have a voice tool, it directly generates and overlays the English audio track into the final video. **v2.1 Fixing "Grime" and "AI Look" (Fixed based on test feedback)**: Four layers of prompts add product clean anchors (clean, pristine, no smudges) and live-action texture anchors (natural skin texture, fabric weaves, handheld feel, natural light). The negative layer excludes stains, grime, wax skin, over-smoothing, plastic feel, and cartoon rendering. Close-up product shots are enlarged and checked shot by shot for cleanliness and realism. If dirty or too fake, anchors are added and re-adjusted, avoiding the delivery of dirty items or wax-like appearances as final work. **If the input image itself is dirty, please switch to a clean and clear product image first**—prompts cannot save a dirty image. **When no generation tools are available in the current environment, the Skill directly states "unable to automatically generate video" and provides a complete prompt package and storyboard table as a fallback delivery**. Regardless of the scenario, it never fakes generation, fakes voiceovers, or fabricates file paths. **It consumes credits**: Every image or video generation call incurs platform credit or subscription quota consumption, subject to the platform's current display. This Skill does not promise specific figures or claim zero cost. Generation results are probabilistic and do not guarantee one-shot success, viral hits, earnings, or platform approval. AI-generated content labeling is handled according to the target platform's current official requirements, with unclear rules marked as "to be verified." Available models and account status depend on your creation interface's real-time display. The Skill first detects tool support, and if unsupported, clearly states so and offers alternatives without faking the use of specified models. ## Three Selling Points 1. **Truly Produces Files**: Delivers storyboard images, video clips, and stitched MP4s, not just a sheet of prompt text. 2. **Reliable English Narration**: Fact anchors -> Graded English scripts -> Audio-video alignment -> Ad compliance -> Mandatory English subtitles + Publishing-layer English voiceover SOP. Honestly explains that the video itself does not speak English and vocals must be added later, avoiding treating background music as narration. 3. **Cures Grime and AI Look (v2.1)**: Clean anchors + Live-action texture anchors + Shot-by-shot enlarged QA ensure close-ups are clean and avoid plastic wax looks. It detects tools, passes 9 gates, reports consumption, and confirms before acting, gracefully falling back if tools are missing.
VideoDrama Storyboard Master
Transforms split short drama script segments into 15-second storyboard prompts ready to be copied into Seedance 2.0 for video generation. Each original number corresponds to an independent video, preserving plot actions, character lines, voiceovers, character relationships, and chronological order without merging, splitting, or reordering the script, making it ideal for continuous mini-dramas and episodic video production. The skill organizes the 15-second shot pacing based on spoken content, detailing shot size, camera movement, facial expressions, gaze, action, atmosphere, lighting, sound effects, and BGM. It distinguishes between dialogue and non-lip-syncing inner monologues, giving dense lines, crowd reactions, emotional payoffs, and key highlights a clear visual focus. The visual descriptions balance generation stability with a drama cinematic feel, adapting to historical border armies, fantasy xianxia, historical political intrigues, modern urban settings, wealthy families, and school themes. Each prompt also includes preceding and succeeding scene details such as positioning, action, and prop status to help adjacent numbers transition naturally. Continuous same-scene plots, scene changes, emotional shifts, eye contact, and audio continuation are matched with appropriate bridging methods. The final output has a clear structure including 15-second time-segment storyboards, unified style terms, and ending scene descriptions for the next video continuation, facilitating direct use in Seedance 2.0.
Information
- Version
- v1
- Last updated
- Runtime credits
- Usage-based
- Models
- Claude Sonnet 4.5