MiniMaxi: MiniMax AI Capabilities via Fhddos Gateway

MiniMaxi gives you access to MiniMax’s full AI capabilities through Fhddos’s unified gateway. All requests use the /minimaxi prefix, and every request body, response field, and error code stays fully compatible with the official MiniMax API — so you can migrate existing integrations without changing a single payload shape.

Field details, value ranges, and constraints follow the official MiniMax API reference. This page covers Fhddos-specific routing, billing, and ready-to-run examples.

Authentication & routing

Set your Fhddos token as a Bearer credential and prefix every path with /minimaxi:

export BASE_URL="https://aiapi.fhddos.com"
export TOKEN="your-fhddos-token"

Every request follows the same pattern:

Authorization: Bearer <fhddos-token>
POST/GET https://aiapi.fhddos.com/minimaxi/<minimax-path>

Capabilities at a glance

Video Generation

Create text-to-video, first-frame, first-last-frame, and subject-reference video tasks with MiniMax-Hailuo-02.

Text-to-Speech (TTS)

Synthesize audio synchronously via t2a_v2 or submit long jobs asynchronously via t2a_async_v2.

Voice Cloning

Clone a voice in three steps — upload audio, optionally upload a prompt sample, then call voice_clone.

File Management

Retrieve file metadata by ID and download raw file content from the MiniMax files API.

Billing

Video generation uses a pay-on-success async billing model:

When you create a task, Fhddos records it but does not immediately charge your account.
A background poller checks each task’s final state (success or failure).
Only successful tasks are billed according to the price schedule.
Failed tasks are never charged.

Video generation

Submit video generation tasks to POST /minimaxi/v1/video_generation. All tasks are asynchronous — creation returns a task_id you then poll to retrieve the finished video.

Scenario 1 — Text-to-video

curl -X POST "$BASE_URL/minimaxi/v1/video_generation" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-Hailuo-02",
    "prompt": "A man picks up a book [Pedestal up], then reads [Static shot].",
    "duration": 6,
    "resolution": "1080P"
  }'

Example response (task created):

{
  "task_id": "106916112212032",
  "platform_id": "video_01JSGXXXXXXXXXXXXXXXXXX",
  "base_resp": {
    "status_code": 0,
    "status_msg": "success"
  }
}

task_id is the raw MiniMax upstream task ID — you can use it directly with the official MiniMax query endpoint. platform_id (video_<ULID>) is Fhddos’s internal tracking ID. Both IDs are accepted by the query endpoint.

Scenario 2 — First-frame image drive

Animate from a single starting image by supplying first_frame_image:

curl -X POST "$BASE_URL/minimaxi/v1/video_generation" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A mouse runs toward the camera, smiling and blinking.",
    "first_frame_image": "https://cdn.hailuoai.com/prod/2024-09-18-16/user/multi_chat_file/9c0b5c14-ee88-4a5b-b503-4f626f018639.jpeg",
    "model": "MiniMax-Hailuo-02",
    "duration": 6,
    "resolution": "1080P"
  }'

Scenario 3 — First-frame + last-frame

Generate the transition between two keyframes by providing both first_frame_image and last_frame_image:

curl -X POST "$BASE_URL/minimaxi/v1/video_generation" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A little girl grows up.",
    "first_frame_image": "https://filecdn.minimax.chat/public/fe9d04da-f60e-444d-a2e0-18ae743add33.jpeg",
    "last_frame_image": "https://filecdn.minimax.chat/public/97b7cd08-764e-4b8b-a7bf-87a0bd898575.jpeg",
    "model": "MiniMax-Hailuo-02",
    "duration": 6,
    "resolution": "1080P"
  }'

Scenario 4 — Subject reference

Pass reference images via subject_reference to anchor a character or object across the generated video:

curl -X POST "$BASE_URL/minimaxi/v1/video_generation" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A girl runs toward the camera and winks with a smile.",
    "subject_reference": [
      {
        "type": "character",
        "image": [
          "https://cdn.hailuoai.com/prod/2025-08-12-17/video_cover/1754990600020238321-411603868533342214-cover.jpg"
        ]
      }
    ],
    "model": "S2V-01"
  }'

Query a video task

Poll task status with the task_id returned at creation:

curl "$BASE_URL/minimaxi/v1/query/video_generation?task_id=$TASK_ID" \
  -H "Authorization: Bearer $TOKEN"

Text-to-speech (TTS)

Fhddos exposes both synchronous and asynchronous TTS endpoints. Choose sync for short, latency-sensitive requests and async for longer texts.

Sync — t2a_v2
Async — t2a_async_v2

Send text and receive the audio file directly in the response:

curl -X POST "$BASE_URL/minimaxi/v1/t2a_v2" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "speech-2.8-hd",
    "text": "Welcome to Fhddos.",
    "output_format": "hex"
  }' \
  --output tts-output.mp3

Submit a job and poll for the result separately:

# Step 1 — submit
curl -X POST "$BASE_URL/minimaxi/v1/t2a_async_v2" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "speech-2.8-hd",
    "text": "This is a longer piece of text synthesized asynchronously."
  }'

# Step 2 — poll
curl "$BASE_URL/minimaxi/v1/query/t2a_async_query_v2?task_id=$TASK_ID" \
  -H "Authorization: Bearer $TOKEN"

Voice aliases

To maintain compatibility with OpenAI-style voice names, Fhddos automatically maps the following aliases to the corresponding MiniMax voice_id values:

Alias	MiniMax voice_id
`alloy`	`female-chengshu`
`echo`	`male-qn-qingse`
`fable`	`male-qn-jingying`
`onyx`	`presenter_male`
`nova`	`presenter_female`
`shimmer`	`audiobook_female_1`

When a voice has a known default emotion, Fhddos automatically fills in emotion if you don’t supply one explicitly. You can override any alias through channel custom parameters:

{
  "voice": {
    "alloy": "female-chengshu|sad",
    "echo": "male-qn-qingse"
  }
}

Key TTS parameters

Parameter	Values	Notes
`output_format`	`hex` (default), `url`	Non-streaming only; URL validity follows upstream policy
`emotion`	`happy`, `sad`, `angry`, `fearful`, `disgusted`, `surprised`, `calm`, `fluent`, `whisper`	Not all model + voice combinations support every emotion
`sound_effects`	`spacious_echo`, `auditorium_echo`, `lofi_telephone`, `robotic`	Applies an acoustic effect to the output

Audio return mode

Control how the audio payload is returned via the audio_mode channel parameter:

Mode	Behaviour
`json` (default)	Response body is JSON; `data.audio` holds the hex string or URL
`hex`	When `output_format=hex`, returns a raw audio stream; when `output_format=url`, still returns JSON

{
  "audio_mode": "hex"
}

Async tasks preserve your original voice_id and parameters in the returned JSON payload. If you used an alias, Fhddos performs the conversion and emotion back-fill before sending the upstream request — not after.

Voice cloning

Clone any voice in three steps using POST /minimaxi/v1/voice_clone. The interface is fully compatible with the official MiniMax /v1/voice_clone endpoint.

Step 1 — Upload the source audio

curl -X POST "$BASE_URL/minimaxi/v1/files/upload" \
  -H "Authorization: Bearer $TOKEN" \
  -F "purpose=voice_clone" \
  -F "file=@/path/to/audio.mp3"

The response includes:

file.file_id — pass this to POST /minimaxi/v1/voice_clone
file.filename — the original file name

Step 2 — (Optional) Upload a prompt sample

Upload a short reference clip to improve similarity and stability:

curl -X POST "$BASE_URL/minimaxi/v1/files/upload" \
  -H "Authorization: Bearer $TOKEN" \
  -F "purpose=prompt_audio" \
  -F "file=@/path/to/short-sample.mp3"

Step 3 — Clone the voice

curl -X POST "$BASE_URL/minimaxi/v1/voice_clone" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "file_id": 1234567890,
    "voice_id": "MiniMax001",
    "text": "In recent years, artificial intelligence has entered a period of rapid development...",
    "model": "speech-2.8-hd",
    "need_noise_reduction": true,
    "need_volume_normalization": true
  }'

Request fields

Field	Required	Description
`file_id`	✅	ID of the audio to clone (from `purpose=voice_clone` upload)
`voice_id`	✅	Target cloned voice ID; must follow official naming rules
`clone_prompt.prompt_audio`	❌	File ID of a short reference sample (from `purpose=prompt_audio` upload)
`clone_prompt.prompt_text`	❌	Transcript of the reference sample
`text`	❌	Preview text to synthesize (max 1 000 characters)
`model`	✅ when `text` set	TTS model for preview synthesis (e.g. `speech-2.8-hd`)
`language_boost`	❌	Minor-language / dialect enhancement (`Chinese`, `English`, `auto`, …)
`need_noise_reduction`	❌	Enable noise reduction on the source audio
`need_volume_normalization`	❌	Normalise volume of the source audio
`aigc_watermark`	❌	Append an AIGC watermark rhythm to the preview audio

Response fields

Field	Description
`demo_audio`	Preview audio URL (returned when `text` + `model` are provided)
`input_sensitive`	Populated when the source audio triggers a content policy; includes `type`
`base_resp.status_code`	`0` = success
`base_resp.status_msg`	Human-readable status description

File management

Retrieve metadata and download content for any file stored in MiniMax’s files API.

Get file metadata

curl "$BASE_URL/minimaxi/v1/files/$FILE_ID" \
  -H "Authorization: Bearer $TOKEN"

Download file content

curl "$BASE_URL/minimaxi/v1/files/$FILE_ID/content" \
  -H "Authorization: Bearer $TOKEN" \
  --output downloaded-file

File IDs are returned by upload endpoints (e.g. voice cloning steps above) and also appear in completed video task responses as download references.

​Authentication & routing

​Capabilities at a glance

Video Generation

Text-to-Speech (TTS)

Voice Cloning

File Management

​Billing

​Video generation

​Scenario 1 — Text-to-video

​Scenario 2 — First-frame image drive

​Scenario 3 — First-frame + last-frame

​Scenario 4 — Subject reference

​Query a video task

​Text-to-speech (TTS)

​Voice aliases

​Key TTS parameters

​Audio return mode

​Voice cloning

​Step 1 — Upload the source audio

​Step 2 — (Optional) Upload a prompt sample

​Step 3 — Clone the voice

​Request fields

​Response fields

​File management

​Get file metadata

​Download file content

Authentication & routing

Capabilities at a glance

Billing

Video generation

Scenario 1 — Text-to-video

Scenario 2 — First-frame image drive

Scenario 3 — First-frame + last-frame

Scenario 4 — Subject reference

Query a video task

Text-to-speech (TTS)

Voice aliases

Key TTS parameters

Audio return mode

Voice cloning

Step 1 — Upload the source audio

Step 2 — (Optional) Upload a prompt sample

Step 3 — Clone the voice

Request fields

Response fields

File management

Get file metadata

Download file content