计费与配额Billing & Quota
平台按模型类型采用不同的计费方式,调用前可在「模型广场」查看各模型价格;每次调用均记录消费明细,可在控制台查询与导出。The platform bills by model type; check each model's price in the Model Square before calling. Every call is logged with usage details, which you can query and export in the console.
计费模式Billing Modes
| 模式Mode | 计算方式Calculation | 典型场景Typical Use |
|---|---|---|
| 按 Token 计费Per Token | 输入 / 输出 / 缓存 token 单价,支持阶梯价Unit price per input / output / cached token, with tiered pricing supported | 文本对话、推理Chat and reasoning |
| 按张计费Per Image | 按生成数量与尺寸 / 质量倍率计费Billed by count and size / quality ratio | 图片生成Image generation |
| 按秒计费Per Second | 单价 × 分辨率倍率 × 实际秒数Unit price × resolution ratio × actual seconds | 视频生成(部分模型)Video generation (some models) |
| 积分计费Per Credit | 按分辨率与时长档位累计积分Credits accumulate by resolution and duration tier | 视频生成(viduq2、kling-v3)Video generation (viduq2, kling-v3) |
| 按字符计费Per Character | 按输入字符数计费Billed by the number of input characters | 语音合成Speech synthesis |
| 按音频时长计费Per Audio Duration | 单价 × 实际音频秒数(不足最低计费时长按最低时长)Unit price × actual audio seconds (billed at the minimum duration when below the minimum billable duration) | 语音转字幕Speech-to-Subtitles (ASR) |
计费方式说明Billing Methods Explained
按 TokenBy Token
输入与输出分别计价,缓存命中使用更低单价;长上下文场景可配置阶梯价。Input and output are priced separately, with a lower unit price for cache hits; tiered pricing can be configured for long-context scenarios.
按张By Image
图片按生成张数计费,尺寸与质量倍率会放大单张价格。Images are billed per generated image; size and quality ratios scale up the per-image price.
按秒By Second
视频按实际生成秒数计费,分辨率倍率参与计算。Videos are billed by actual generated seconds, with the resolution ratio factored in.
积分By Credit
按分辨率与时长档位累计积分,再按每积分单价折算金额。Credits accumulate by resolution and duration tier, then convert to an amount at the per-credit price.
按字符By Character
语音合成按输入字符数计费。Speech synthesis is billed by the number of input characters.
按音频时长By Audio Duration
语音转字幕按上游识别出的音频实际时长(秒)计费,提交时按最低计费时长预检余额,完成后按真实时长结算。Speech-to-Subtitles (ASR) is billed by the actual audio duration (seconds) recognized upstream. Balance is pre-checked against the minimum billable duration at submission, and settlement uses the actual duration on completion.
积分计费Credit-Based Billing
积分档位按分辨率与是否带视频输入区分,由模型价格设置中的「积分档位」配置;积分按分辨率与时长档位累计后,再按每积分单价折算金额并应用账户折扣。Credit tiers differ by resolution and whether video input is present, configured as "Credit Tiers" in the model price settings. Credits accumulate by resolution and duration tier, then convert to an amount at the per-credit price, with the account discount applied.
余额与充值Balance & Top-Up
| 充值方式Top-Up Method | 说明Description |
|---|---|
| 在线充值Online Top-Up | 在控制台「钱包」页面选择金额完成支付,到账后立即可用。Choose an amount on the Wallet page in the console to complete payment; it becomes available immediately. |
| 兑换码Redemption Code | 在「兑换」页面输入兑换码,充值对应额度到账户。Enter a redemption code on the Redeem page to add the corresponding quota to your account. |
| 管理员充值Admin Top-Up | 联系平台管理员为账户充值或调整余额。Contact a platform administrator to top up your account or adjust the balance. |
- 余额用于所有模型调用,调用成功后在消费明细中展示计费金额。Balance covers all model calls; the billed amount appears in the usage details after a successful call.
- 账户折扣在计费时自动应用,无需手动计算。The account discount is applied automatically during billing; no manual calculation needed.
- 余额不足时请求会被拒绝,充值后即可恢复调用。Requests are rejected when the balance is insufficient; calls resume after topping up.
消费记录Usage Records
- 每次调用记录消费日志:时间、模型、调用方式、计费明细与金额。Every call is logged with: time, model, calling method, billing details, and amount.
- 控制台支持按时间 / 模型 / 类型筛选,并支持导出。The console supports filtering by time / model / type, and supports export.
- 日志默认至少保留 30 天。Logs are retained for at least 30 days by default.
语音转字幕Speech-to-Subtitles (ASR)
语音转字幕基于火山 AI MediaKit 异步任务接口:提交音 / 视频地址后返回任务 ID,任务完成后返回音频时长与逐句字幕(含起止时间)。平台同步落库任务并按音频实际时长计费(audio_second),任务失败不计费。Speech-to-Subtitles (ASR) is built on the Volcano AI MediaKit async task API: submit an audio / video URL to get a task ID; on completion it returns the audio duration and per-sentence subtitles (with start/end times). The platform persists the task and bills by the actual audio duration (audio_second); failed tasks are not billed.
提交任务Submit the Task
携带令牌请求 POST /v1/tools/asr-subtitles,Body 传 model(计费模型名,如 asr-subtitles)与 video_url 或 audio_url(公网可访问的音 / 视频地址),同步返回 task_id。Call POST /v1/tools/asr-subtitles with your token. The body carries model (the billing model name, e.g. asr-subtitles) and video_url or audio_url (a publicly accessible audio / video URL); the response returns task_id synchronously.
轮询结果Poll the Result
请求 GET /v1/tools/asr-subtitles/{task_id} 查询状态:running(进行中)、completed(已完成)、failed(失败);结果 24 小时内有效,建议拿到终态后尽快取走。Call GET /v1/tools/asr-subtitles/{task_id} to check status: running (in progress), completed (done), or failed. Results are valid for 24 hours — retrieve them promptly once a terminal state is reached.
取走字幕并结算Retrieve Subtitles and Settle
completed 时返回 result.duration(音频秒数)与 result.subtitles 字幕数组;平台按 duration × 每秒单价 结算并记录消费明细。When completed, the response returns result.duration (audio seconds) and the result.subtitles array; the platform settles by duration × 每秒单价 and records the usage details.
| 请求字段Request Field | 类型Type | 必填Required | 说明Description |
|---|---|---|---|
model | string | 建议必填Recommended | 计费模型名(如 asr-subtitles),缺省时按 asr-subtitles 计费Billing model name (e.g. asr-subtitles); defaults to billing as asr-subtitles when omitted |
video_url / audio_url | string | 至少一个At least one | 待识别的公网视频 / 音频地址Public video / audio URL to be transcribed |
language / enable_speaker_info / enable_confidence / client_token / callback_url | - | 否No | 原样透传上游,详见 AI MediaKit 官方文档Passed through to the upstream as-is; see the official AI MediaKit documentation |
curl -X POST "https://your-gateway/v1/tools/asr-subtitles" \
-H "Authorization: Bearer sk-your-token" \
-H "Content-Type: application/json" \
-d '{"model":"asr-subtitles","video_url":"https://example.com/demo.mp4"}'
# → {"success":true,"task_id":"amk-tool-asr-subtitles-xxxx","request_id":"..."}
查询示例:Query example:curl "https://your-gateway/v1/tools/asr-subtitles/amk-tool-asr-subtitles-xxxx" \
-H "Authorization: Bearer sk-your-token"
# → {"task_id":"...","status":"completed",
# "result":{"duration":15.092,"subtitles":[{"start_time":0,"end_time":3150,"text":"..."}]}}
常见问题FAQ
| 问题Question | 解答Answer |
|---|---|
| 为什么扣费与标价看起来不一致?Why does the charged amount look different from the listed price? | 计费可能受上下文阶梯、尺寸 / 分辨率档位等因素影响,以消费明细中的实际金额为准。Billing may be affected by context tiers, size / resolution tiers, and other factors; the actual amount in the usage details prevails. |
| 如何查看我的折扣?How do I check my discount? | 联系平台管理员确认账户折扣;计费时折扣已自动应用在最终金额中。Contact a platform administrator to confirm your account discount; it is applied automatically to the final amount at billing time. |
| 视频生成为什么比预估贵?Why is video generation more expensive than estimated? | 视频按实际生成秒数与分辨率档位计费,最终以消费明细中的实际金额为准。Videos are billed by actual generated seconds and resolution tier; the actual amount in the usage details prevails. |
| 消费明细能导出吗?Can usage details be exported? | 可以。在日志 / 消费明细页面筛选后即可导出,日志默认至少保留 30 天。Yes. Filter on the logs / usage details page and export; logs are retained for at least 30 days by default. |
下一步Next Steps
- 视频生成Video Generation — 查看积分计费模型的实际调用实例。 — See real call examples for credit-billed models.
- API 参考API Reference — 查看各接口的计费维度与请求示例。 — See billing dimensions and request examples for each endpoint.