MYAI168 Chat API Guide

This page shows how to call the site chat endpoint for each use case: text chat, image recognition, image generation, speech recognition, speech generation and video generation. The module lists and curl examples in each section are generated from the modules currently enabled on this site; modules that are not enabled are not shown.

Every use case calls the same endpoint; only module and its parameters change:

POST https://www.myai168.com/en/ai/chat/chat

Authentication

Authenticate with your developer API key in the Authorization: Bearer YOUR_API_KEY header (the X-Dev-Key header is also accepted). Replace YOUR_API_KEY in the examples with your own key. Get a key at: https://www.myai168.com/en/ai/user/developer

Request format

Send the request as a form: use application/x-www-form-urlencoded for text-only requests (curl --data-urlencode) and multipart/form-data when uploading files (curl -F, with @ before the file name). Add -N to curl so the stream prints in real time.

On this site the total size of one request, including all uploaded files, is limited to 512 MB; larger requests are rejected with HTTP 413.

Common parameters

Image and video generation modules accept additional fields (for example quality, aspect_ratio, resolution, duration) described in their sections; fields a module does not use are ignored.

Module codes available on this site

Show all 116 module codes

Module code	Name
search	Search
default	Site Default Model
agent	AI Agent
gemini_4_argon	Google Gemini 4 Argon
gemini_3_8_flash	Google Gemini 3.8 Flash
gemini_3_8_live	Live Voice Conversation gemini-3.8-live
gemini_3_8_live_extended_thinking	Live Voice Conversation gemini-3.8-live-extended-thinking
gemini_3_6_flash	Google Gemini 3.6 Flash
gemini_3_5_flash_lite	Google Gemini 3.5 Flash Lite
gemini_3_1_pro	Google Gemini 3.1 Pro
gemini_3_1_flash_live	Live Voice Conversation gemini-3.1-flash-live
gemini_3_1_flash_lite	Google Gemini 3.1 Flash Lite
claude_fable_5_1	Anthropic Claude Fable 5.1
claude_fable_5	Anthropic Claude Fable 5
claude_opus_5_5	Anthropic Claude Opus 5.5
claude_opus_5	Anthropic Claude Opus 5
claude_opus_4_8	Anthropic Claude Opus 4.8
claude_sonnet_5_5	Anthropic Claude Sonnet 5.5
claude_sonnet_5	Anthropic Claude Sonnet 5
claude_haiku_5_5	Anthropic Claude Haiku 5.5
claude_haiku_4_5	Anthropic Claude Haiku 4.5
gpt_6_astra	OpenAI GPT-6 Astra
gpt_6_1_sol	OpenAI GPT-6.1 Sol
gpt_6_sol	OpenAI GPT-6 Sol
gpt_6_luna	OpenAI GPT-6 Luna
openai_decisions	OpenAI Decisions
gpt_5_6_sol	OpenAI GPT-5.6 Sol
gpt_5_6_terra	OpenAI GPT-5.6 Terra
gpt_5_6_luna	OpenAI GPT-5.6 Luna
chatgpt_5_5	OpenAI ChatGPT 5.5
chatgpt_5_4_mini	OpenAI ChatGPT 5.4 mini
chatgpt_5_4_nano	OpenAI ChatGPT 5.4 nano
perplexity_sonar_deep_research	Perplexity Sonar Deep Research
perplexity_sonar_reasoning_pro	Perplexity Sonar Reasoning Pro
perplexity_sonar_pro	Perplexity Sonar Pro
grok_4_7	xAI Grok 4.7
grok_4_6	xAI Grok 4.6
grok_4_5	xAI Grok 4.5
muse_spark_1.3	Meta Muse Spark 1.3
kimi_k3	Kimi K3
glm_5_2	GLM 5.2
qwen3.5_397b_a17b	Qwen 3.5 397B A17B
knowledge_chunk	Knowledge Chunk Splitter
meeting_minutes	Meeting Minutes
notes	Note Organizer
slide_outline	Slide Outline
slide2	Slide Generation 2
slide	Slide Generation
editor	AI Copywriter / Copywriting
official_document	Official Document Writing
chatpdf	Chat with PDF (Text Only)
homework	Homework Assistant
social_media_content_improver	Social Post Optimization
contract	Contract Review
report	Report Generation
webpage_summary	Webpage Summary
google_search	Google Search
google_news	Google News
paper	Paper Writing
speech_to_text	Speech Recognition
gpt_transcribe	Speech Recognition gpt_transcribe
whisper_1	Speech Recognition whisper_1
gpt_4o_transcribe_diarize	Multi-Speaker Speech Recognition
programming	Programming Assistant
plan	Project Plan Writing
news_report	Press Release
news_oral_report	Oral News Broadcast Script
ocr	Image OCR
document_ocr_to_md	Document OCR to Markdown
document_ocr_to_docx	Document OCR to DOCX
gai_article_detector	AI Content Detector
summary_500	500-Word Summary
summary_1000	1000-Word Summary
writing_optimizer	Writing Optimization
essay_grading	Essay Grading
zenrows	Web Crawler
get_text	Document Text Extraction
gpt_image_2_5_flare	GPT Image 2.5 Flare Generation
gpt_image_2_5_sunburst	GPT Image 2.5 Sunburst Generation
gemini_nano_banana_2_1	Nano Banana 2.1
nano_banana_pro	Nano Banana Pro
nano_banana_3_1_flash	Nano Banana 3.1 Flash
image_to_prompt	Image-to-Prompt
gemini_omni_flash_1_1	Gemini Omni Flash 1.1 Video Generation
music	Music Generation
music_extend	Extend Song
music_cover	Song Restyle
music_voice	Sing with Your Voice
music_stems	Vocal & Instrumental Separation
change_clothes	AI Change Clothes
seedream_5_0	BytePlus Seedream 5.0 Pro Text-to-Image
seedance_2_5	seedance 2.5 Generate Video
seedance_2_0	seedance 2.0 Generate Video
cheer_video	Cheer Video Generation
christmas_video	Christmas Video Generation
rich_new_year_video	New Year Video Generation
gemini_3_8_flash_tts	Google Gemini 3.8 Flash Text to Speech
gemini_3_8_flash_lite_tts	Google Gemini 3.8 Flash-Lite Text to Speech
gemini_2_5_flash_tts	Google Gemini 2.5 Flash Text to Speech
gemini_2_5_pro_tts	Google Gemini 2.5 Pro Text to Speech
text_to_speech_byteplus	BytePlus Seed-TTS
seedaudio_1_0	BytePlus SeedAudio 1.0 Audio Generation
voice_replication	BytePlus Voice Cloning
translate_to_tc	Translate to Traditional Chinese
translate_to_english	Translate to English
english_gramma	English Grammar Check
speak_female	Multilingual Voice Tutor (Female) - Sophie
speak_male	Multilingual Voice Tutor (Male) - Alex
travel_translator	Travel Translator
gemini_3.5_live_translate	Live Voice Translation gemini-3.5-live-translate
gpt_live_transcribe	Live Transcription gpt_live_transcribe
sc2tc	Simplified to Traditional
tc2sc	Traditional to Simplified
simplified_chinese_detector	Simplified Chinese Detector
check_typo	Typo Checker
health_education	Health Education

Billing and errors

Every successful call is billed to your account credits according to the model pricing, and the reply ends with a usage summary; failed or interrupted calls are never charged. Errors are returned as text inside the stream with HTTP 200 (for example an invalid key, insufficient credits or a module that is not enabled), not as HTTP error codes; a request larger than the size limit is rejected with HTTP 413.

None of these cases is charged. Suggested handling: if the stream ends without any reply text or output file, treat the content of the html events as the failure reason; the wording follows the site language, so do not match whole sentences exactly.

Questions, writing, translation, programming and other text work all use text chat: put a text model code in module and your question in input; the reply is streamed piece by piece.

default means the default model of this site. To use a specific model, replace it with any code in the “Module codes available on this site” list of the quick start section.

Basic conversation

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=default" \
  --data-urlencode "input=Hello, please introduce yourself."

Search the web before answering

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=default" \
  --data-urlencode "search_internet=true" \
  --data-urlencode "input=Summarize today's top technology news."

search_internet only takes effect on modules that support web search; each search is billed.

Continue a conversation (session_sn)

On the first call, omit session_sn. Read the data: {"session_sn": N} event from the stream and save the value, then send it back on every follow-up call of the same conversation.

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=default" \
  --data-urlencode "input=Recommend three books about time management."

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=default" \
  --data-urlencode "session_sn=YOUR_SESSION_SN" \
  --data-urlencode "input=Which one is best for beginners?"

Replace YOUR_SESSION_SN with the value returned by the first call.

Reasoning effort (effort)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=claude_opus_5_5" \
  --data-urlencode "effort=high" \
  --data-urlencode "input=Prove that the square root of 2 is irrational."

effort sets how hard a reasoning model thinks (common values: low, medium, high; some models also accept minimal, xhigh, max). Accepted values depend on the model; the model default applies when omitted. display set to summarized appends a summary of the thinking process to the reply on models that support it.

Decisions (predicate, choice, score)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=openai_decisions" \
  --data-urlencode "param2=choice" \
  --data-urlencode "param3=Which department should handle this complaint?" \
  --data-urlencode "param4=billing|Payments, invoices and refunds
technical|Problems using the product
shipping|Delivery and tracking
other|Anything else" \
  --data-urlencode "input=I was charged twice for my order."

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=openai_decisions" \
  -F "param2=predicate" \
  -F "param3=Does this photo show a damaged product?" \
  -F "upload_file[]=@photo.jpg"

param2 is the question type (predicate, choice, score), param3 is the question or condition, and param4 lists the options (choice) or levels (score) one per line as value|description. The reply is a table with the probability of each option and a confidence value. Upload images with upload_file[] to evaluate pictures; when param3 is omitted, the input text is used as the question and the images as the evidence.

Upload an image and ask about it in input; a module that accepts images answers based on the picture (describing it, reading text, analysing charts and so on). Upload images in the upload_file[] field of a multipart/form-data request.

Ask a question about an image

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=default" \
  -F "input=What is in this picture? Please describe it in detail." \
  -F "upload_file[]=@photo.jpg"

Repeat upload_file[] to send several images in one call. Accepted formats, image count and size limits depend on the module.

Extract the text in an image (OCR)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=ocr" \
  -F "input=Extract all the text in this image." \
  -F "upload_file[]=@scan.png"

C# example (ocr.cs)

using System;
using System.IO;
using System.Net.Http;
using System.Net.Http.Headers;
using System.Text;
using System.Text.Json;
using System.Threading.Tasks;

class Program
{
    static async Task Main()
    {
        string url = "https://www.myai168.com/en/ai/chat/chat";
        string imagePath = "scan.png";

        using var http = new HttpClient { Timeout = TimeSpan.FromMinutes(10) };
        http.DefaultRequestHeaders.Authorization =
            new AuthenticationHeaderValue("Bearer", "YOUR_API_KEY");

        using var form = new MultipartFormDataContent();
        form.Add(new StringContent("ocr"), "module");
        form.Add(new StringContent("Extract all the text in this image."), "input");
        form.Add(new ByteArrayContent(await File.ReadAllBytesAsync(imagePath)),
                 "upload_file[]", Path.GetFileName(imagePath));

        using var request = new HttpRequestMessage(HttpMethod.Post, url) { Content = form };
        using var response = await http.SendAsync(
            request, HttpCompletionOption.ResponseHeadersRead);
        response.EnsureSuccessStatusCode();

        var text = new StringBuilder();
        var notice = new StringBuilder();
        using var reader = new StreamReader(
            await response.Content.ReadAsStreamAsync(), Encoding.UTF8);
        string? line;
        while ((line = await reader.ReadLineAsync()) != null)
        {
            if (!line.StartsWith("data: "))
            {
                continue;
            }
            using var doc = JsonDocument.Parse(line.Substring(6));
            JsonElement root = doc.RootElement;
            if (root.TryGetProperty("html", out JsonElement html))
            {
                notice.AppendLine(html.GetString());
                continue;
            }
            if (!root.TryGetProperty("choices", out JsonElement choices)
                || choices.GetArrayLength() == 0)
            {
                continue;
            }
            JsonElement choice = choices[0];
            if (choice.TryGetProperty("delta", out JsonElement delta)
                && delta.TryGetProperty("content", out JsonElement content)
                && content.ValueKind == JsonValueKind.String)
            {
                text.Append(content.GetString());
            }
            if (choice.TryGetProperty("finish_reason", out JsonElement finish)
                && finish.ValueKind == JsonValueKind.String
                && finish.GetString() == "stop")
            {
                break;
            }
        }

        Console.OutputEncoding = Encoding.UTF8;
        Console.WriteLine(text.Length > 0 ? text.ToString() : notice.ToString());
    }
}

Requires .NET 6 or later. The example uploads scan.png with MultipartFormDataContent, reads the stream line by line and joins the text in choices[0].delta.content; when the stream ends without any text, the content of the html events is the reason for the failure. Keep the file extension in the uploaded file name, because the image format is determined from it.

Modules on this site that accept images

Describe the picture you want in input to generate an image; upload a reference image to edit it, or send back session_sn to keep editing the previous result. Generated images are returned as download links (see “Reading the response and getting output files”).

Generate an image from text

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gpt_image_2_5_flare" \
  --data-urlencode "input=A cat sitting on a desk, digital illustration"

Edit with a reference image

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=gpt_image_2_5_flare" \
  -F "input=Change the background to a beach at sunset" \
  -F "upload_file[]=@photo.png"

Keep editing the previous image (session_sn)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gpt_image_2_5_flare" \
  --data-urlencode "session_sn=YOUR_SESSION_SN" \
  --data-urlencode "input=Make the cat wear a red scarf"

Image modules that support continuous editing emit data: {"session_sn": N} on the first call; send it back and the previous result is used as the source image when no new image is uploaded.

Choose the image quality (quality)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gpt_image_2_5_flare" \
  --data-urlencode "quality=high" \
  --data-urlencode "input=A watercolor painting of a lighthouse at dawn"

Image parameters

Image generation modules on this site

Upload an audio file to turn speech into text; you can set the language, and some modules can output SRT subtitles.

Transcribe an audio file

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=speech_to_text" \
  -F "param2=en" \
  -F "upload_file[]=@speech.mp3"

input is not needed: speech recognition modules do not use typed text.

Output SRT subtitles

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=speech_to_text" \
  -F "param2=en" \
  -F "param3=on" \
  -F "upload_file[]=@speech.mp3"

Help the module spell names and terms

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=speech_to_text" \
  -F "param2=en" \
  -F "hint=Lin Mei-ling, TSMC, Kubernetes" \
  -F "upload_file[]=@speech.mp3"

Speech recognition parameters

Speech recognition modules on this site

Turn the text in input into speech, choosing the voice and language with param2 and related parameters; the output is a download link to an MP3 file.

Text to speech (Google Gemini 3.8 Flash Text to Speech)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gemini_3_8_flash_tts" \
  --data-urlencode "param2=Kore" \
  --data-urlencode "param3=en-US" \
  --data-urlencode "param4=Warm and cheerful" \
  --data-urlencode "input=Welcome to our website. We hope you have a wonderful day."

The same parameters apply to gemini_3_8_flash_tts, gemini_3_8_flash_lite_tts, gemini_2_5_flash_tts, gemini_2_5_pro_tts.

Voice names (30)

Value	Description
Zephyr	Bright
Puck	Upbeat
Charon	Informative
Kore	Firm
Fenrir	Excited
Leda	Young
Orus	Firm
Aoede	Breezy
Callirrhoe	Easygoing
Autonoe	Bright
Enceladus	Breathy
Iapetus	Clear
Umbriel	Easygoing
Algieba	Smooth
Despina	Smooth
Erinome	Clear
Algenib	Gravelly
Rasalgethi	Informative
Laomedeia	Upbeat
Achernar	Soft
Alnilam	Firm
Schedar	Even
Gacrux	Warm
Pulcherrima	Forward
Achird	Friendly
Zubenelgenubi	Casual
Vindemiatrix	Gentle
Sadachbia	Lively
Sadaltager	Knowledgeable
Sulafat	Warm

Language codes (30)

Value	Description
de-DE	German (Germany)
en-AU	English (Australia)
en-GB	English (United Kingdom)
en-IN	English (India)
en-US	English (United States)
es-US	Spanish (United States)
fr-FR	French (France)
hi-IN	Hindi
pt-BR	Portuguese (Brazil)
ar-XA	Arabic
es-ES	Spanish (Spain)
fr-CA	French (Canada)
id-ID	Indonesian (Indonesia)
it-IT	Italian (Italy)
ja-JP	Japanese (Japan)
tr-TR	Turkish (Turkey)
vi-VN	Vietnamese (Vietnam)
bn-IN	Bengali (India)
gu-IN	Gujarati (India)
kn-IN	Kannada (India)
ml-IN	Malayalam (India)
mr-IN	Marathi (India)
ta-IN	Tamil (India)
te-IN	Telugu (India)
nl-NL	Dutch (Netherlands)
ko-KR	Korean (South Korea)
cmn-CN	Chinese
pl-PL	Polish (Poland)
ru-RU	Russian (Russia)
th-TH	Thai (Thailand)

Text to speech (BytePlus Seed-TTS)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=text_to_speech_byteplus" \
  --data-urlencode "param2=zh_female_wanwanxiaohe_moon_bigtts" \
  --data-urlencode "input=Welcome to our website. We hope you have a wonderful day."

Voice ids (33)

Value	Description
zh_female_wanwanxiaohe_moon_bigtts	Isabella - Sweet Taiwanese
zh_female_cancan_mars_bigtts	Luna - Bright All-Purpose Female
zh_male_baqiqingshu_mars_bigtts	Edward - Bold Mature Male
zh_female_wenroushunv_mars_bigtts	Emma - Gentle Lady
zh_female_shaoergushi_mars_bigtts	Tina - Children's Stories
zh_male_silang_mars_bigtts	William - Magnetic Male
zh_male_jieshuonansheng_mars_bigtts	James - Narration Male
zh_female_jitangmeimei_mars_bigtts	Grace - Warmhearted Girl
zh_female_tiexinnvsheng_mars_bigtts	Sophia - Caring Female
zh_female_qiaopinvsheng_mars_bigtts	Mia - Playful Female
zh_female_mengyatou_mars_bigtts	Ava - Cute Girl
zh_female_qingxinnvsheng_mars_bigtts	Olivia - Fresh Female
zh_female_linjia_mars_bigtts	Lily - Girl Next Door
en_female_anna_mars_bigtts	Anna - Gentle American
en_male_adam_mars_bigtts	Adam - Clear American
en_female_sarah_mars_bigtts	Sarah - Australian Female
en_male_dryw_mars_bigtts	Dryw - Deep Australian
en_male_smith_mars_bigtts	Smith - Deep British
zh_male_guozhoudege_moon_bigtts	Andrew - Cantonese Male
zh_female_daimengchuanmei_moon_bigtts	Elena - Sichuan Cute Girl
zh_male_jingqiangkanye_moon_bigtts	Thomas - Beijing Chatter
zh_female_wanqudashu_moon_bigtts	Robert - Playful Uncle
zh_female_gaolengyujie_moon_bigtts	Charlotte - Cool Elegant Female
zh_male_wennuanahu_moon_bigtts	Mark - Warm Male
zh_female_linjianvhai_moon_bigtts	Lila - Fresh Girl Next Door
zh_male_shaonianzixin_moon_bigtts	Ethan - Confident Youth
zh_male_yuanboxiaoshu_moon_bigtts	Joseph - Knowledgeable Male
zh_male_yangguangqingnian_moon_bigtts	George - Sunny Young Male
zh_female_shuangkuaisisi_moon_bigtts	Aria - Lively Female
multi_female_shuangkuaisisi_moon_bigtts	Haruko - Energetic Female (JA/ES)
multi_male_jingqiangkanye_moon_bigtts	Kazune - Male (JA/ES)
multi_female_gaolengyujie_moon_bigtts	Akemi - Cool Female (JA)
multi_male_wanqudashu_moon_bigtts	Hiroshi - Mature Male (JA/ES)

Read text in a cloned voice (BytePlus Voice Cloning)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=voice_replication" \
  -F "input=Welcome to our website. We hope you have a wonderful day." \
  -F "upload_file[]=@voice_sample.mp3"

Upload a clear recording of the voice to imitate (upload_file[]) and put the text to read in input.

Generate sound from a description (BytePlus SeedAudio 1.0 Audio Generation)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=seedaudio_1_0" \
  --data-urlencode "input=Gentle rain on a window with distant thunder, calm atmosphere"

Describe the sound effects, voice or atmosphere in input. Optionally attach up to 3 reference audio clips (cite them in the text as @Audio1, @Audio2 …) or one reference image; audio and images cannot be mixed.

Speech generation modules on this site

Describe the scene in input to generate a video; you can set the aspect ratio, resolution and length, or upload images as the first frame and reference materials. The output is a download link to an MP4 file.

Generate a video from text (seedance 2.5 Generate Video)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=seedance_2_5" \
  --data-urlencode "aspect_ratio=16:9" \
  --data-urlencode "resolution=1080p" \
  --data-urlencode "duration=10" \
  --data-urlencode "generate_audio=true" \
  --data-urlencode "input=A cat jumping over a fence in slow motion, cinematic lighting"

Set the first and last frames

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=seedance_2_5" \
  -F "input=The flower slowly blooms" \
  -F "first_frame=@start.jpg" \
  -F "last_frame=@end.jpg"

The first/last frames cannot be combined with reference materials (keyframes[] or upload_file[]) in the same call; such a call is rejected before generation and not charged.

Generate a video from text (Gemini Omni Flash 1.1 Video Generation)

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gemini_omni_flash_1_1" \
  --data-urlencode "resolution=1080p" \
  --data-urlencode "aspect_ratio=9:16" \
  --data-urlencode "input=A cat jumping over a fence in slow motion, cinematic lighting"

Video generation usually takes a few minutes. The stream keeps sending progress messages; keep the connection open until the final event (an interrupted call is not charged).

Video generation modules on this site

Call a GPTs custom assistant you created, or a public one on this site, through the site-native chat endpoint: send module=gpts and pass the assistant id in param2. The id is the number at the end of its chat page URL (https://www.myai168.com/en/ai/gpts/gpts?sn=YOUR_GPT_SN).

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gpts" \
  --data-urlencode "param2=YOUR_GPT_SN" \
  --data-urlencode "input=Hello, please introduce yourself."

Continue the same conversation — replace YOUR_SESSION_SN with the session_sn value emitted in the previous reply stream:

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=gpts" \
  --data-urlencode "param2=YOUR_GPT_SN" \
  --data-urlencode "session_sn=YOUR_SESSION_SN" \
  --data-urlencode "input=Please continue."

The assistant role instructions and bound knowledge bases are applied automatically — no need to repeat them in input. To continue the same conversation, send back the session_sn value emitted in the stream. Billing follows this site default model; a private assistant can only be called with its creator API key.

AI Agent plans a task as several steps and works through them round by round with tools (web search, reading web pages and documents, writing documents, producing charts and files) until the task is done. Send module=agent and describe the task in input. Each new conversation creates a notebook of its own; uploaded files and output files are stored in that notebook.

Run a task

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=agent" \
  --data-urlencode "input=Hello, please introduce yourself."

Continue a conversation (session_sn)

On the first call, omit session_sn. The stream sends a data: {"session_sn": N} event as soon as the task starts; save the value and send it back on every follow-up call of the same conversation (follow-up calls do not send a new number). AI Agent then remembers the earlier rounds and keeps working in the same notebook, so files uploaded or produced earlier are still there and can be revised.

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=agent" \
  --data-urlencode "input=Research the electric scooter market and write a short report as a Word document."

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  --data-urlencode "module=agent" \
  --data-urlencode "session_sn=YOUR_SESSION_SN" \
  --data-urlencode "input=Change the conclusion of the report into a bulleted list."

Replace YOUR_SESSION_SN with the value returned by the first call.

When a task needs many steps, AI Agent may stop and ask whether to continue; reply “Continue” with the same session_sn and it carries on from where it stopped. To start an unrelated task, omit session_sn and a new notebook is created.

Hand over files

curl -N https://www.myai168.com/en/ai/chat/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "module=agent" \
  -F "input=Read this document and list five key points." \
  -F "upload_file[]=@report.pdf"

Uploaded files (upload_file[], multiple files allowed) are saved in the notebook of the conversation for AI Agent to read and edit.

AI Agent parameters

Values of chat_model (25)

Value	Name
default	Site Default
gemini_4_argon	Gemini 4 Argon
gemini_3_8_flash	Gemini 3.8 Flash
gemini_3_6_flash	Gemini 3.6 Flash
gemini_3_5_flash_lite	Gemini 3.5 Flash Lite
gemini_3_1_pro	Gemini 3.1 Pro
gemini_3_1_flash_lite	Gemini 3.1 Flash Lite
claude_fable_5_1	Claude Fable 5.1
claude_fable_5	Claude Fable 5
claude_opus_5_5	Claude Opus 5.5
claude_opus_5	Claude Opus 5
claude_opus_4_8	Claude Opus 4.8
claude_sonnet_5_5	Claude Sonnet 5.5
claude_sonnet_5	Claude Sonnet 5
claude_haiku_5_5	Claude Haiku 5.5
claude_haiku_4_5	Claude Haiku 4.5
gpt_6_1_sol	GPT-6.1 Sol
gpt_6_sol	GPT-6 Sol
gpt_6_luna	GPT-6 Luna
gpt_5_6_sol	GPT-5.6 Sol
gpt_5_6_terra	GPT-5.6 Terra
gpt_5_6_luna	GPT-5.6 Luna
chatgpt_5_5	ChatGPT 5.5
chatgpt_5_4_mini	ChatGPT 5.4 mini
chatgpt_5_4_nano	ChatGPT 5.4 nano

Reading the reply

The reply of AI Agent arrives as html events that also carry an md key with the same content in Markdown; join the md values in order to get the full reply. html events without an md key show the progress of each round (steps and tool actions) and the credit summary at the end, and can be skipped. System messages, such as insufficient credits or the question whether to continue, arrive as text chunks in choices[0].delta.content.

Download links of output files appear in the reply, in the form https://www.myai168.com/en/api/file/download_nb?sn=...&fn=.... They are private files of the calling account: sign in to this site with the same account and open them in a browser, or open the notebook of the conversation on the AI Agent Notebook page.

Python example (two calls in the same conversation)

import json
import requests

url = "https://www.myai168.com/en/ai/chat/chat"
headers = {"Authorization": "Bearer YOUR_API_KEY"}


def ask(text, session_sn=0):
    data = {"module": "agent", "input": text}
    if session_sn:
        data["session_sn"] = session_sn
    reply = []
    with requests.post(url, headers=headers, data=data, stream=True, timeout=(10, 1800)) as r:
        r.raise_for_status()
        r.encoding = "utf-8"
        for line in r.iter_lines(decode_unicode=True):
            if not line or not line.startswith("data: "):
                continue
            payload = json.loads(line[6:])
            if "session_sn" in payload:
                session_sn = payload["session_sn"]
            elif "md" in payload:
                reply.append(payload["md"])
            elif payload.get("choices"):
                choice = payload["choices"][0]
                content = (choice.get("delta") or {}).get("content")
                if content:
                    reply.append(content)
                if choice.get("finish_reason") == "stop":
                    break
    return "".join(reply), session_sn


answer, session_sn = ask("Research the electric scooter market and write a short report as a Word document.")
print(answer)
print("session_sn:", session_sn)

answer, session_sn = ask("Change the conclusion of the report into a bulleted list.", session_sn)
print(answer)

Requires pip install requests. ask() returns the reply and the conversation number; the second call sends the number back to continue the same conversation.

Things to note

• Your account needs at least 10,000 credits to use AI Agent. A task runs many rounds of reasoning and tool actions and can use a large amount of credits, so consider setting a credit limit on the API key.
• A task can take several minutes or longer and is complete only when finish_reason is stop, so set a long read timeout on the client. If the connection drops, the task stops and the rounds already completed are still charged by usage.
• Only one task can run in the same notebook at a time. Wait until the previous call has finished before sending the next one; otherwise the reply says the previous task is still running.
• session_sn only works for the account that started the conversation, so use an API key of the same account. AI Agent conversations are not listed on the history page; keep session_sn in your program.

Response events

The response is an SSE stream (Content-Type: text/event-stream, UTF-8 encoded); each event is one data: {JSON} line:

Skip any key not listed above — more event kinds may be added later.

Usage summary of the call

A successful call sends one html event before the stream ends, wrapped in <span class="billing-summary">, stating the credits used by the call, the elapsed time and the remaining credits. The wording follows the site language and suits showing to users as is; when a program needs numbers, query the balance with GET https://www.myai168.com/en/api/user/get_credits.

Getting output files

Images, videos, audio and other outputs appear as download URLs of this site in the href / src attributes of the html fragments (URLs look like https://www.myai168.com/en/api/file/download_photo?id=... or https://www.myai168.com/en/api/file/download?id=...). These files are private to the calling account: sign in to this site with the same account and open the URL in a browser to download them.

Python example

The example reads the stream, assembles the full text reply and collects the conversation id, the output file URLs and the usage summary. If the stream ends with no reply text and no output file, the notices received are treated as the failure reason (install the package first: pip install requests):

import json
import re
import requests

url = "https://www.myai168.com/en/ai/chat/chat"
headers = {"Authorization": "Bearer YOUR_API_KEY"}
data = {"module": "default", "input": "Hello, please introduce yourself."}

text_parts = []
file_urls = []
notices = []
billing = ""
session_sn = 0

with requests.post(url, headers=headers, data=data, stream=True, timeout=(10, 900)) as r:
    r.raise_for_status()
    r.encoding = "utf-8"
    for line in r.iter_lines(decode_unicode=True):
        if not line or not line.startswith("data: "):
            continue
        payload = json.loads(line[6:])
        if "session_sn" in payload:
            session_sn = payload["session_sn"]
        elif "html" in payload:
            fragment = payload["html"]
            for u in re.findall(r'(?:href|src)="([^"]*/api/file/[^"]+)"', fragment):
                if u not in file_urls:
                    file_urls.append(u)
            plain = " ".join(re.sub(r"<[^>]+>", " ", fragment).split())
            if 'class="billing-summary"' in fragment:
                billing = plain
            elif plain:
                notices.append(plain)
        elif payload.get("choices"):
            choice = payload["choices"][0]
            content = (choice.get("delta") or {}).get("content")
            if content:
                text_parts.append(content)
            if choice.get("finish_reason") == "stop":
                break

answer = "".join(text_parts)
if not answer and not file_urls:
    raise RuntimeError("request failed: " + " / ".join(notices))
print(answer)
print("session_sn:", session_sn)
print("files:", file_urls)
print("billing:", billing)

Do not omit r.encoding = "utf-8": the response header does not state the encoding, and without this line requests turns non-ASCII text into garbage. timeout=(10, 900) means 10 seconds to connect and at most 900 seconds between two chunks.

To upload files, pass them in the files argument with the field name upload_file[]:

files = [("upload_file[]", open("photo.jpg", "rb"))]
data = {"module": "default", "input": "What is in this picture?"}
r = requests.post(url, headers=headers, data=data, files=files, stream=True, timeout=(10, 900))

Node.js example

This uses the fetch built into Node.js 18 and later, so no package is needed. Save it as chat.mjs and run node chat.mjs:

const url = "https://www.myai168.com/en/ai/chat/chat";
const body = new URLSearchParams({ module: "default", input: "Hello, please introduce yourself." });

const res = await fetch(url, {
  method: "POST",
  headers: { Authorization: "Bearer YOUR_API_KEY" },
  body,
});
if (!res.ok) throw new Error("HTTP " + res.status);

let answer = "";
let sessionSn = 0;
const fileUrls = new Set();
const decoder = new TextDecoder();
let buffer = "";

for await (const chunk of res.body) {
  buffer += decoder.decode(chunk, { stream: true });
  const lines = buffer.split("\n");
  buffer = lines.pop();
  for (const line of lines) {
    if (!line.startsWith("data: ")) continue;
    const payload = JSON.parse(line.slice(6));
    if (payload.session_sn !== undefined) {
      sessionSn = payload.session_sn;
    } else if (payload.html !== undefined) {
      for (const m of payload.html.matchAll(/(?:href|src)="([^"]*\/api\/file\/[^"]+)"/g)) {
        fileUrls.add(m[1]);
      }
    } else if (payload.choices) {
      answer += payload.choices[0].delta?.content ?? "";
    }
  }
}

console.log(answer);
console.log("session_sn:", sessionSn);
console.log("files:", [...fileUrls]);

To upload files, use FormData (do not set Content-Type yourself — fetch adds the boundary automatically); read the response the same way as above:

import { readFile } from "node:fs/promises";

const form = new FormData();
form.append("module", "default");
form.append("input", "What is in this picture?");
form.append("upload_file[]", new Blob([await readFile("photo.jpg")]), "photo.jpg");

const res = await fetch("https://www.myai168.com/en/ai/chat/chat", {
  method: "POST",
  headers: { Authorization: "Bearer YOUR_API_KEY" },
  body: form,
});

Practical tips

• A call is complete only when finish_reason is stop. A call interrupted midway is not charged and can simply be sent again.
• Waiting time varies a lot by use: text usually takes seconds to tens of seconds, image and speech generation can exceed a minute, and video generation can take several minutes. Give the client a generous read timeout.
• Keep a separate session_sn for each conversation, and omit it when starting a new topic.
• Keep the key in server-side code only, never in a web page or a mobile app; setting a credit limit on each key is recommended.